Effect of TV Drama Piracy

Information & Management 59 (2022) 103599
Contents lists available at ScienceDirect
Information & Management

journal homepage: www.elsevier.com/locate/im
The effect of TV drama piracy: An analysis of digital piracy users, internet

buzz, and TV drama viewership
Dongyeon Kim a, Kyuhong Park b, Youngsok Bang c, *
a
Department of Business Administration, Catholic University of Korea, 43 Jibong-ro, Bucheon, Gyeonggi 14662, Republic of Korea
b
Department of Business Administration, Inha University, 100 Inha-ro, Michuhol-gu, Incheon 22212, Republic of Korea
c
School of Business, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul 03722, Republic of Korea
A R T I C L E I N F O A B S T R A C T
Keywords: Notwithstanding the biggest victim of digital piracy, we have a limited understanding of how TV drama piracy
Digital piracy affects its legal consumption. Unlike movies or music, TV dramas are typically a series of episodes, where pirated
BitTorrent past episodes may not necessarily carry a negative impact on the legal consumption of current episodes. The
Internet buzz
consumption boosted by piracy can promote Internet buzz, which may draw the attention of potential consumers
TV viewership
Mediation
and may result in broader legal consumption afterward. Based on an empirical analysis regarding the re
Restricted maximum likelihood estimation lationships among the viewership of 20 TV dramas broadcast in the US, social media activities on Twitter, Google
and YouTube search trends, and the number of BitTorrent users who pirated these TV dramas, we show that
although the direct effect of media consumption through BitTorrent to TV viewership is negative, the indirect
effect that mediates Internet buzz is positive, which dilutes the negative effect of digital piracy. Such positive
effects of digital piracy are more substantial for the second half of dramas, less popular dramas, and serial
dramas. We discuss the theoretical contributions and managerial implications of our study.
1. Introduction word-of-mouth, the consumption boosted by digital piracy can promote

Internet buzz, which may draw the attention of potential consumers,
Digital piracy has been regarded as the most critical issue to address thereby resulting in increased legal consumption of subsequent epi
in the media industry. The characteristics of information goods that are sodes. Thus, the overall effect of digital piracy on TV dramas would
easily copied and transmitted through the Internet make them vulner depend on the relative size of the substitution effect, where consumers in a
able to digital piracy. Several efforts have been devoted to preventing or piracy channel would choose a legal channel if the piracy channel were
discouraging digital piracy by using technological solutions such as not available, and the attention effect, where the Internet buzz promoted
Digital Rights Management (DRM), pricing strategies such as versioning, by piracy users elicits the attention of potential consumers, leading them
and anti-Internet piracy laws on copyright enforcement and manage to legal consumption.
ment [19,41]. Statistics show, however, that digital piracy is still prev Previous business cases also indicate the possible coexistence of
alent: users made 174 billion visits to Internet piracy sites in 2019 [37]. substitution and attention effects of digital piracy. Black Widow suffered
TV shows and dramas are reported as the biggest victims, taking up a sharp decline in theatrical revenues, with its high-quality pirate
49.4% of all digital piracy activities [4]. version released in its second week [31]. However, in the Pirate Bay
As a critical threat to digital content businesses, many studies have shutdown in Sweden, we noted a drop in word-of-mouth and a
empirically investigated the effects of movie piracy (e.g., [7,15,26,28, contemporaneous decline in ticket sales [33]. Furthermore, ShotX could
30,32,38,43]) or music piracy (e.g., [1,3,22,24,35]). Notwithstanding achieve a noticeable boost in sales as its free torrent release garnered
the biggest victims of digital piracy, we have paid little attention to its wide attention from potential game users [46]. David Petrarca, the di
effects on TV shows or dramas. Unlike movies or music, TV dramas are rector of Game of Thrones, also mentioned that he was not bothered
typically a series of episodes, where pirated past episodes may not about Game of Thrones’ rampant piracy because it might boost a cultural
necessarily carry a negative impact on the legal consumption of current buzz [45].
episodes. Specifically, when media consumption is subject to As such, we investigate the effects of digital piracy on TV dramas,
* Corresponding author.
E-mail address: [email protected] (Y. Bang).
https://doi.org/10.1016/j.im.2022.103599
Received 25 August 2020; Received in revised form 11 January 2022; Accepted 17 January 2022
Available online 19 January 2022
0378-7206/© 2022 Elsevier B.V. All rights reserved.
D. Kim et al. Information & Management 59 (2022) 103599
with a particular focus on the mediating role of Internet buzz in the Table 1
relationship between media consumption in a piracy channel and a legal Effects of digital piracy on sales.
channel. The mediation effect may be heterogeneous, depending on Title Data/Context Findings
various factors, as well. For example, the effect could be larger for a less
Insignificant Effects on Sales
popular drama because it may require more public awareness. Under Andersen and Microeconomic survey data in No association exists
standing the heterogeneous effects is critical for developing appropriate Frenz [3] Canadian P2P file sharing and between the number of P2P
counter-piracy strategies. Thus, we also consider such moderating fac CD purchasing behavior files downloaded and CD
tors in our analysis. (music album) album sales.
Aguiar and Clickstream of 16,500 No evidence of a decrease in
To examine the mediating role of Internet buzz, we collect data, Martens [1] European consumers in 2011 digital music sales by
including the viewership history of 20 TV dramas broadcast in the US, unlicensed downloads.
social media activities on Twitter regarding those TV dramas, Google Martikainen BitTorrent (movie file sharing) The file-sharing elasticity of
and YouTube search trends of those dramas, and the number of Bit [30] Nash Information Services sales is not statistically
DVD sales in 2009 different from zero.
Torrent users who pirated those dramas during their broadcast periods.
McKenzie [32] Box office revenue for 6083 There is not compelling
Our empirical analysis shows that although the direct effect of media movies between 2005 and evidence of revenue increase
consumption through BitTorrent on TV viewership was negative, the 2013 in 16 countries by piracy prevention
indirect effect mediating Internet buzz was positive, which diluted the policies.
negative effect of digital piracy. Consequently, the overall piracy effect Oberholzer-Gee OpenNap Music (file sharing) File sharing has no
and Strumpf Nielsen SoundScan (music significant effect on
on TV viewership was not significant. [35] album sales) in 2002 purchases of the average
This attention effect can be strong enough to neutralize the substi album.
tution effect, especially when Internet buzz drives the media consump Smith and BitTorrent (movie file sharing) The availability of pirated
tion of potential legal consumers. The attention effect would prevail as Telang [43] Amazon DVD sales on copies of movies has no
2005–2006 impact on subsequent DVD
Internet users continue to generate additional buzz. We confirm our
sales.
conjecture with further analysis, which replicates the main analysis for Negative Effects on Sales
the first and second halves of the broadcasting period of these TV Danaher and Movie rental data 2011–2013 The Antipiracy Act
dramas, considering that there is more buzz concerning the dramas as Smith [7] (shutdown of piracy sites)
subsequent episodes are progressively broadcast. The analytical results results in a revenue increase
of 6.5~8.5%.
from the first half reveal minimal mediation effects across the TV
Godinho de Subscription video-on-demand Treated households (free
dramas, whereas the results from the second half show significant Matos et al. (SVoD) services and BitTorrent SVoD offered) increase TV
mediation effects that could even result in an overall positive effect of [12] usage consumption and decrease
digital piracy. Internet use for both illegal
downloads and uploads.
Also, the attention effect can be more substantial for less popular
Hennig-Thurau Longitudinal panel survey in File sharing destroys a
dramas because they need greater public awareness. We replicate our et al. [15] 2006, German piracy and significant amount of movie
analysis for dramas with high and low viewership and find that the movie purchases industry revenues (9.4%
attention effect was significant only for low viewership dramas. Partic decrease in total industry
ularly, the positive effect was greater for their second half. In high revenue).
Liebowitz [24] U.S. Census population survey, File sharing leads to a
viewership dramas, the attention effect was not significant both in their
DMA’s population data, significant decline in record
first and second halves. This finding is consistent with prior movie pi Nielsen Media research data sales.
racy studies, revealing promotional effects of piracy for movies in need (music)
of public awareness (e.g., [21,38]). Lastly, we compare the model esti Liu [26] Torrent, box office and DVD Piracy reduces the total
sales, and Google trends in revenue of motion pictures
mates between procedural dramas (that do not need prior viewing of any
2015 by 2.71%, but the number is
other episode to understand, e.g., ‘CSI’) and serial dramas (that require smaller than expected.
viewing previous episodes, e.g., ‘Lost’). In serial dramas, the story is Ma et al. [28] Box office revenue (movie) and Pre-release piracy reduces
interwoven across episodes; thus, consumers would be highly motivated the dates of pirated copy box office revenues by 19%.
to watch the next episode by the Internet buzz about past episodes. The postings
Positive or Conditional Effects on Sales
positive word-of-mouth effect was significant only for such serial
Koschmann and Box office revenue between High-quality illegal copies
dramas. Qian [21] 2013 and 2014 (173 movies) may increase box office
The rest of the paper is organized as follows. Next, we review the and Torrent data from The revenues at launch but
prior literature on the effects of digital piracy and Internet buzz. Then, Pirate Bay decrease the revenues
afterward.
we describe our data and empirical approach. Subsequently, we present
Kretschmer and Music sales between 2009 and The blockage of YouTube
the results of our analysis. Lastly, we conclude with a discussion of the Peukert [22] 2013 music videos may decrease
results and implications of the study. music sales.
Lu et al. [27] Digital piracy (Torrent data The presence of post-release
2. Related Literature from The Pirate Bay), Word-of- piracy is associated with a
Mouth (Data from Rotten 3% increase in box office
Tomatoes), and Box office revenue.
This study builds on and contributes to two streams of research. The revenue (movie)
first stream empirically examines the effect of digital piracy on legal Ma et al. (2016) Box office revenues between The elimination of piracy
consumption across different contexts. The second stream focuses on the [29] 2006 and 2008 (533 movies) may increase 15% box office
revenue on average. Pre-
role of Internet buzz in generating consumer demand.
release piracy, however, may
increase the revenue for
2.1. The effect of digital piracy some movies.
Peukert et al. Weekly box office revenues of Piracy site shutdown may
The effects of digital piracy on the legal consumption of information [38] 308 movies across 14 countries increase box office revenues
for movies in the top decile
goods have drawn a great deal of attention from the Information Systems of the number of opening
community during the last decade. A plethora of empirical studies have screens but decrease for
examined whether illegal movie or music consumption through a file- movies in the bottom decile.
sharing platform or an application (e.g., BitTorrent) indeed affects Zhang [49] Music sold between January Removing DRM increases
1992 and June 2012 digital music sales by 10%.
2
legitimate consumption. Despite the negative public stance against movies in that a series of episodes are broadcast over time. Regardless of
digital piracy, prior studies have yielded mixed empirical findings piracy or legal channels, those who consume early episodes may
regarding how digital content consumption in a piracy channel affects generate Internet buzz, which may draw the attention of potential
consumption in a legal channel. Table 1 summarizes the results of pre consumers and may result in broader legal consumption afterward. Such
vious studies. a word-of-mouth effect could prevail for late episodes, given that the
Several studies provide empirical evidence revealing a negative ef Internet buzz would snowball and make the drama more visible on the
fect of digital piracy on content sales. For example, based on a longi Internet over time. Further, different drama characteristics may play a
tudinal panel survey, Hennig-Thurau et al. [15] argue that illegal media role in determining the piracy effects, which we identify as moderators
consumption through file-sharing platforms decreased revenues of the for the piracy effect.
movie industry by 9.4%. Likewise, Liebowitz [24] shows that illegal
music consumption through file-sharing platforms resulted in a large 2.2. Internet buzz and consumer demand
decline in record sales of the US music industry. Similarly, Ma et al. [28]
show that pre-released piracy copies of a new film reduced box office The effect of Internet buzz on consumer demand has been empiri
revenue by 19%. Danaher and Smith [7] further investigate the causal cally examined across different contexts, including online communities,
impact of the shutdown of piracy sites and find that the Antipiracy Act blogs, Twitter, and Google trends. Prior studies have consistently
increased the revenue of movie rentals by 6.5%–8.5%. Lastly, Godinho revealed that Internet buzz can work as a sales indicator or positively
de Matos et al. [12] show that a free subscription to video-on-demand affect future sales.
(SVoD) services for a certain period can increase households’ TV con For example, based on an analysis of conversations on a popular
sumption and decrease their illegal uploads and downloads in the piracy online community site, Godes and Mayzlin [11] provide evidence that
channel after the complimentary subscription period, suggesting a online buzz about TV shows significantly increases their TV ratings.
negative association between legal and illegal media consumption. Dhar and Chang [9] also show that the volume of blog posts and future
Contrary to the abovementioned empirical findings, another large sales of music albums are positively correlated. Asur and Huberman [2]
body of literature suggests an insignificant effect of digital piracy. For used Twitter data to forecast the box-office revenues of movies. They
example, Oberholzer-Gee and Strumpf [35] examine the relationship argue that a prediction model based on tweets provides better accuracy
between illegal consumption on file-sharing platforms and music album than the Hollywood Stock Exchange, a well-established wisdom-of-
sales, revealing an insignificant effect. Similarly, Smith and Telang [43] crowds prediction platform. Echoing this notion, Rui et al. [40] examine
show that the availability of pirated movie copies has no impact on how word-of-mouth on Twitter affects movie sales based on a dynamic
subsequent DVD sales. In addition, based on microeconomic survey data panel model. They show that Twitter users with more followers affect
in Canada, Andersen and Frenz [3] find an insignificant relationship sales to a greater extent than users with fewer followers, thus implying
between the number of P2P file downloads and CD album sales. Marti the positive effect of social media buzz on sales. Correspondingly,
kainen [30] focuses on the relationship between movie sharing through Hennig-Thurau et al. [16] analyze the impact of word-of-mouth on
BitTorrent and DVD sales, and shows that the ’file-sharing elasticity of Twitter regarding consumers’ adoption of new movies; they find a sig
sales’ is not statistically different from zero. nificant effect of word-of-mouth, especially for tweets with negative
There have been attempts to explain the above insignificant effects. sentiment. Dewan and Ramaprasad [8] focus on the interplay among
For example, the instrument variable used in Oberholzer-Gee and blog buzz, radio play, and music sales, and find a stronger buzz effect on
Strumpf [35] has been criticized regarding its criteria [25]. Smith and sales for niche music relative to mainstream music. Lastly, a recent
Telang [43] also raise concerns about the data collection period related empirical study focuses on Google search trends and finds that they
to an insignificant effect. However, a systematic explanation as to why positively affect market response [10].
the piracy effect could be negligible was not provided. Given that Internet buzz involving TV dramas (such as social media
Recent empirical studies provide evidence of positive or conditional postings or search trends) is promoted by users who consume or will
effects of digital piracy, where they focus on its promotional effects. In consume these media contents, the piracy users who would not watch
this regard, Lu et al. [27] explain why digital piracy could increase sales the drama if the piracy channel were not available could contribute to
with four different mechanisms: trial, network effects, advertising, and the Internet buzz and, subsequently, the TV viewership as well. Recent
word-of-mouth. As such, efforts to reduce piracy may bring unexpected statistics illustrate that 90% of piracy activities are conducted outside
results. For example, Zhang [49] shows that removing DRM could in the US, mostly in developing countries [4]. That is because of lenient
crease digital music sales by approximately 10%. Kretschmer and Peu anti-piracy laws or actions in such countries, which lessen piracy users’
kert [22] also show that the blockage of YouTube music videos may psychological burden of punishment [36]. Further, such users might
decrease music sales. Furthermore, piracy activities can work as an perceive their piracy as a victimless crime for their own country, as
advertiser particularly when the digital content is unnoticed by the popular content is generated from developed countries [14].
public. For example, Peukert et al. [38] reveal that piracy site shutdown TV viewership is estimated from the US residents only, whereas non-
may increase box office revenues for movies in the top decile of the US residents can also make Internet buzz. As such, those piracy users
number of opening screens but decrease for movies in the bottom decile. outside the US would not bring a significant substitution effect as they
Koschmann and Qian [21] also find that high-quality illegal copies may are (although they may promote piracy activities in the US.) Thus, pi
increase box office revenues at launch, then decrease the revenues af racy users on BitTorrent might not necessarily impose a negative effect
terward. Lastly, based on the correlation between the presence of digital but could positively affect TV viewership in the US.
piracy and the volume of IMDb reviews, Lu et al. [27] argue that movie Although we provide our rough predictions, how media consumption
piracy may create word-of-mouth, thereby counteracting the substitu in the piracy channel affects media consumption in the legitimate
tion effect of digital piracy. channel remains an empirical question, which would depend on the
Although we have rich empirical evidence regarding the effects of relative sizes of the substitution effect and the attention effect. We
movie or music piracy, we still have a limited empirical understanding empirically examine these effects in the next section.
of the effects of TV drama piracy. TV dramas are different from music or
3
Table 2 Table 4
Data sources. Descriptive statistics.
Data Source Details Variable Mean Std. Dev. Min Max
Drama TV by the TV ratings, Number of viewers TV_rating 0.881 0.391 0.2 2.2
viewership Numbers Viewers (millions) 4.657 2.919 0.540 12.520
Digital piracy µTorrent Number of BitTorrent users (Numbers of Seeds 1381.039 1303.796 0 8409
seeders and leechers) Leeches 333.466 451.856 0 4966
Internet buzz Google, Google trends, YouTube trends Google 11.765 13.309 0 100
YouTube YouTube 12.226 14.483 0 100
Twitter Numbers of tweets, replies, retweets, likes, Tweets 718.242 3025.247 0 79,335
and quotes Engagements 6739.040 30,716.28 0 649,292
Drama quality IMDb IMDb star ratings, Award Interactions 7457.281 33,306.28 2 674,440
Table 3 leechers who own only a part of the file so that they upload that part but
Program list. download the other part at the same time (the variable is named
’Leeches’). For our main analysis, we use the total number of BitTorrent
Title Broadcaster Broadcasting date
users who were involved in the piracy of those dramas, i.e., the sum of
Arrow The CW 3/8, 3/29, 4/5, 4/12, 4/19, 4/26, 5/3, 5/10, 5/ the numbers of seeders and leechers (the variable is named ’Piracy’). We
17
also investigate the separate piracy effects of seeders and leechers.
Black Lightning The CW 3/6, 3/13, 3/20, 3/27, 4/3, 4/10, 4/17
Bull CBS 3/6, 3/13, 3/27, 4/3, 4/17, 5/1, 5/8 Third, we observed the Google and YouTube trends of those dramas
Criminal Minds CBS 3/7, 3/14, 3/21, 3/28, 4/4, 4/11, 4/18 during the same periods, which involve the number of titles of those
Gotham FOX 3/8, 3/15, 3/22, 3/29, 4/5, 4/12, 4/19, 5/3, 5/ dramas that are used as the search terms on those platforms. As such,
10, 5/17
those trends reflect the current attention of Internet users. Both the
Hawaii Five-0 CBS 3/9, 3/30, 4/6, 4/13, 4/20, 4/27, 5/4, 5/11, 5/
18 Google and YouTube search trends are exposed to general Internet users,
Jane the Virgin The CW 3/9, 3/16, 3/23, 4/6, 4/13, 4/20 which may entice people’s attention to popular keywords, including TV
Life in Pieces CBS 3/8, 3/29, 4/5, 4/12, 4/19, 4/26, 5/3, 5/17 drama titles. We use both Google and YouTube search trends to inves
Lucifer FOX 3/12, 3/19, 3/26, 4/16, 4/23, 4/30, 5/7, 5/14, tigate their mediating role in the relationship between media con
5/28
sumption through BitTorrent and TV viewership (the variables are
Man with a CBS 3/5, 3/12, 3/19, 3/26, 4/9, 4/16, 4/30, 5/7, 5/
Plan 14, 5/21 named ’Google’ and ’YouTube,’ respectively).
Modern Family ABC 3/7, 3/21, 3/28, 4/4, 4/11, 5/2, 5/9, 5/16 Fourth, we obtained data from AC Nielsen with respect to social
Scandal ABC 3/8, 3/15, 3/29, 4/5, 4/12, 4/19 media users’ activities on Twitter regarding those dramas during the
Scorpion CBS 3/5, 3/19, 3/26, 4/9, 4/16
same periods. The data include the number of tweets that mentioned the
SEAL Team CBS 3/7, 3/21, 3/28, 4/11, 4/25, 5/2, 5/9, 5/16
Supernatural The CW 3/8, 3/29, 4/5, 4/12, 4/19, 4/26, 5/3, 5/10, 5/
title, actors’ names/handles, or character names, or that had fan-
17 generated hashtags or phrases of those dramas. The data also include
Superstore NBC 3/8, 3/15, 3/22, 3/29, 4/5, 4/12, 4/19, 4/26, 5/ the number of activities related to those tweets, such as replies, retweets,
3 likes, and quotes. We adopt three different measures of social media
The Blacklist NBC 3/7, 3/14, 4/4, 4/11, 4/25, 5/2, 5/9, 5/16
activity for our analysis: the number of tweets, engagements, and in
The Flash The CW 3/6, 3/13, 4/10, 4/17, 4/24, 5/1, 5/8, 5/15, 5/
22 teractions regarding those dramas (the variables are named ’Tweets,’
The Goldbergs ABC 3/7, 3/21, 3/28, 4/4, 4/11, 5/2, 5/9, 5/16 ’Engagements,’ and ’Interactions,’ respectively). The number of engage
Young Sheldon CBS 3/8, 3/29, 4/5, 4/12, 4/19, 4/26, 5/3, 5/10 ments is defined as being equal to the sum of the numbers of replies,
retweets, likes, and quotes; the number of interactions is defined as
being equal to the sum of the numbers of engagements and tweets. Both
3. Empirical Approach
measures for engagement and interaction are defined by Twitter and are
widely used in academia (e.g., [18,47]) or in practice (e.g., [6,34]).
3.1. Data
Those social media activities are expected to play a similar role as
Internet search trends, given that they can work as advertisements for
We obtained data from multiple sources for our analysis (Table 2).
Twitter users, such as followers.
First, we gathered data concerning the viewership of 20 TV dramas
Lastly, we collected the IMDb star ratings of each episode and
broadcast in the US from early March 2018 to early June 2018 (Table 3).
whether to receive the Primetime Emmy Award as a quality measure for
Specifically, we collected the data of their TV ratings and the estimated
the episode.2 We must control for the quality of the episode to obtain
number of viewers over time from the TV by the Numbers website,1
correct estimates because the higher quality would simultaneously
which are used as our main dependent variables (the variables are
attract greater digital piracy, more Internet searches and social media
named ’TV_rating’ and ’Viewers,’ respectively).
activities, and higher viewership, all of which could generate spurious
Second, we crawled µTorrent, the most popular BitTorrent client
relationships among our variables of interest.
owned by BitTorrent itself, throughout the broadcasting periods of those
Our longitudinal sample is a panel of 20 dramas with their unbal
TV dramas to collect the number of piracy users for each episode. Spe
anced broadcasting periods, for up to 90 days.3 Table 4 illustrates the
cifically, we collected the number of seeders who distribute each episode
of those dramas on the platform twice a day at 12 AM and 12 PM PST/
PDT, and we use the daily average number of those for our analysis (the
2
variable is named ’Seeds’). Similarly, we collected the number of IMDb star ratings are employed as a proxy for content quality in several
prior studies (e.g., [5, 20, 44]). A Primetime Emmy Award honors the best in U.
S. prime time television programming. It is typically awarded to a group of
episodes (e.g., drama season).
1 3
The website is widely recognized as a trustworthy source of statistics on TV There are five public broadcasters in the U.S. (except PBS, which is
viewership in the U.S. and is cited by well-established media outlets such as education-oriented), which broadcast thirty-six dramas in total during our
CNN, the Associated Press, National Public Radio, and the former sister pub observation window. Among them, twenty dramas remain after we exclude
lication of the Chicago Tribune (https://en.wikipedia.org/wiki/TV_by_th dramas with less than five episodes, dramas suspended for a month and then
e_Numbers). broadcast twice a day, and one outlier which was extremely popular.
4
Table 5
Correlations.
(1) (2) (3) (4) (5) (6) (7) (8) (9)
TV_rating 1.000
Viewers 0.819 1.000
Seeds − 0.314 − 0.371 1.000
Leeches − 0.161 − 0.183 0.726 1.000
Google − 0.247 − 0.112 0.285 0.150 1.000
YouTube − 0.309 − 0.187 0.503 0.281 0.833 1.000
Tweets − 0.088 − 0.130 0.181 0.237 0.133 0.100 1.000
Engagements − 0.104 − 0.136 0.199 0.250 0.137 0.122 0.843 1.000
Interactions − 0.104 − 0.137 0.200 0.252 0.138 0.122 0.868 0.999 1.000
descriptive statistics of our focal variables used in the analysis. The Buzzi,t = α2 + βD2 Digital Piracyi,t + γ 2 Controli,t + εi,t (Eq. 2)
average TV ratings and the average estimated number of viewers of all
episodes are 0.88% and 4.66 million, respectively. The most popular TV Viewershipi,t = α3 + βD3 Digital Piracyi,t− 1 + βB3 Buzzi,t−
(Eq. 3)
1
episode achieved a rating of 2.2%, which corresponds to 12.5 million
+ γ 3 Controli,t + εi,t ,
viewers. The number of seeders (leechers) had an average of 1381 (333)
and a maximum of 8,409 (4,966). The average Google and YouTube
where TV_Viwershipi,t is the TV rating or the estimated number of viewers
search trend measures were 11.8 and 12.2, respectively. Some episodes
of drama i at t, Digital_Piracyi,t-1 is the number of BitTorrent users that
reached a measure of 100, which is the highest possible value of those
pirate drama i at t-1, Buzzi,t is the Google or YouTube trend measure of
measures. The daily number of tweets was 718.2 on average and 79,335
drama i at t, Controli,t is a vector of the control variables of drama i at t, α1
at the maximum. The average numbers of engagements and interactions
(α2 and α3) is the intercept in the first (second and third) equation, βD1
were 6,739 and 7,457, respectively. Note that the average number of
(βD2 and βD3) is the coefficient of Digital_Piracyi,t-1 in the first (second and
interactions equals the sum of the average number of tweets and en
third) equation, βB3 is the coefficient of Buzzi,t-1 in the third equation, γ 1
gagements, following its definition.
(γ 2 and γ 3) is a vector of the coefficients for Controli,t in the first (second
Table 5 provides the correlations between these variables. TV_rating
and third) equation, and εi,t is the disturbance term.
and Viewers are highly, but not perfectly correlated (0.819). Google and
An empirical concern regarding the estimation of our model is
YouTube are highly correlated (0.833), as expected. Tweets is also highly
correlated with both Engagements (0.843) and Interactions (0.868). En
gagements and Interactions are almost perfectly correlated (0.999). Table 6
Analysis results with search trends.
3.2. Empirical model (a) Relationships among Viewers, (b) Relationships among TV_rating,
Google, and Piracy Google, and Piracy
Following our predictions, the digital piracy of TV dramas, measured DV: Viewers DV: TV_rating
by the number of piracy users on BitTorrent, could not only adversely Piracy − 0.021 (0.001)* Piracy − 0.010 (0.015)
Controls Yes Controls Yes
affect TV viewership (substitution effect), but could also promote DV: Google DV: Google
Internet buzz on those TV dramas, which may attract the attention of Piracy 0.051 (0.016)** Piracy 0.051 (0.016)**
potential viewers and may enhance TV viewership (attention effect). To Controls Yes Controls Yes
tease out the substitution and attention effects, we use a mediation DV: Viewers DV: TV_rating
Google 0.104 (0.018)*** Google 0.166 (0.028)***
model adapted from Krull and MacKinnon [23]. Given our multilevel
Piracy − 0.025 (0.010)** Piracy − 0.019 (0.015)
data structure (in our study, each drama has more than or equal to five Controls Yes Controls Yes
observations), the ordinary least squares estimator would underestimate Indirect effect 0.005 Indirect effect 0.008
standard errors of our parameters due to the common errors shared by Direct effect − 0.026 Direct effect − 0.019
episodes within the same drama. Their multilevel mediation technique Total effect − 0.020 Total effect − 0.010
Observations 1113 Observations 1113
allows us to estimate error terms for both episodes (lower-level units)
and dramas (higher-level units) by explicitly modeling the within-group (c) Relationships among Viewers, (d) Relationships among TV_rating,
homogeneity of errors. YouTube, and Piracy YouTube, and Piracy
Our empirical model consists of three equations: 1) TV viewership DV: Viewers DV: TV_rating
Piracy − 0.021 (0.010)* Piracy − 0.010 (0.015)
such as TV ratings or the estimated number of viewers (the DV) on the
number of piracy users (the IV); 2) the Google (or YouTube) search DV: YouTube DV: YouTube
trends or social media activities (the MV) on the number of piracy users Piracy 0.052 (0.018)** Piracy 0.052 (0.018)**
(the IV); and 3) TV viewership measures (the DV) on the Google (or Controls Yes Controls Yes
YouTube) search trends or social media activities (the MV) and the DV: Viewers DV: TV_rating
YouTube 0.035 (0.017)* YouTube 0.087 (0.026)***
number of piracy users (the IV). The direct, indirect, and total effects are Piracy − 0.022 (0.010)* Piracy − 0.015 (0.015)
calculated by the estimation results of the three equations. We also Controls Yes Controls Yes
include dummies indicating each drama (e.g., Modern Family), each Indirect effect 0.002 Indirect effect 0.005
broadcaster (e.g., NBC), each week (e.g., the first week of the observa Direct effect − 0.022 Direct effect − 0.015
Total effect − 0.020 Total effect − 0.010
tion), each day of the week (e.g., Monday), and the broadcast day of the
week for each drama (e.g., every Thursday for Grey’s Anatomy) in our
models to control for their fixed effects. We also control for the quality of Note: †p < 0.1.
*
each episode with IMDb star ratings and dummies indicating whether to p < 0.05.
**
p < 0.01.
receive the award, and the length of each episode in minutes. Our model ***
p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control
specifications are:
variables include the IMDb star rating and runtime of each episode, and
TV Viewershipi,t = α1 + βD1 Digital Piracyi,t− 1 + γ 1 Controli,t + εi,t (Eq. 1) dummies indicating the dramas, day of the week, broadcasting day, broad
casting week, broadcaster, and award.
5
Table 6(a) shows the estimates of the relationships among Viewers,

Google, and Piracy. The estimated number of viewers of drama i at t is
negatively associated with the average number of BitTorrent users who
pirate drama i during t-1 (βD1 = − 0.021, p < 0.05), suggesting a negative
piracy effect on TV viewership. However, this effect is counteracted by
the mediating effect of Internet buzz (Google), where the average num
ber of users who pirate drama i during t is positively associated with the
Google trend (βD2 = 0.051, p < 0.01), and the Google trend at t-1 is also
positively associated with TV viewership at t (βB3 = 0.104, p < 0.001).
As a result, more than 20% of the negative direct effect of piracy (Direct
effect = − 0.026) was offset by the mediating effect (Indirect effect =
0.005). Table 6(b) shows even stronger mediating effects for TV ratings,
where more than 40% of the negative effect of piracy on TV ratings
(Direct effect = − 0.019) was offset by the mediating effect (Indirect
effect = 0.008). Tables 6(c) and 6(d) reveal consistent results for the
YouTube trend.
Fig. 1 visualizes the direct and indirect effects. The positive medi
ating effect of the Internet buzz measure (Google or YouTube) was found
for both TV viewership measures (Viewers and TV_rating), confirming our
conjecture that Internet buzz may play a mediating role between media
consumption in piracy and legitimate channels.
Similarly, we confirm the mediating role of Internet buzz with
Twitter activities. Tables 7(a), (b), (c), (d), (e), and (f) illustrate the
analysis results. Table 7(a) illustrates the results for Tweets. The negative
piracy effect on TV viewership is diluted by the mediating effect of
Tweets, where the average number of users who pirate drama i during t is
positively associated with Tweets (βD2 = 0.071, p < 0.001), and Tweets at
t-1 is also positively associated with TV viewership at t (βB3 = 0.083, p <
0.001). As a result, more than 20% of the negative direct effect of piracy
effect = 0.006). Table 7(b) shows stronger mediating effects for TV
ratings, where almost 50% of the negative effect of piracy on TV ratings
effect = 0.009). Tables 7(c), (d), (e), and (f) reveal consistent results for
Engagements and Interactions, respectively.
Fig. 2 visualizes the direct and indirect effects. The positive medi
Fig. 1. Results with search trends.
ating effect of Twitter measures (Tweets, Engagements, or Interactions)
was found for both TV viewership measures (Viewers and TV_rating),
which weakens the negative substitution effects on TV viewership that
possible reverse causality from Buzz to Digital_Piracy. We conduct a
directly come from digital piracy.
panel vector autoregression (PVAR) model among our variables of in
terest, considering their endogenous nature (Appendix 1). The results
3.4. Additional analysis
consistently indicate that the path from Digital_Piracy to Buzz is signifi
cant, whereas the path from Buzz to Digital_Piracy is insignificant. Our
The mediating effect could be stronger for late versus early episodes
PVAR analysis also confirms the optimal lag length of one for Digital_
of the same drama because the Internet buzz would snowball and make
Piracy. As such, the first episode of each drama is excluded from our
the drama more visible on the Internet over time. We split the entire
analysis.
sample into two subsamples, the first and second halves of the episodes
We estimate the parameters of the three equations based on the
of each drama, and we replicate the analysis for each subsample (Ta
restricted maximum likelihood (REML), which is preferred to the
bles 8, 9, 10, and 11). The mediating effect was negligible for episodes in
traditional maximum likelihood (ML) estimator in the estimation of
the first half, such that the negative effect of digital piracy prevailed
variance components. For our path analysis (the estimation of the direct
during the first half. The mediating effect of Internet buzz was signifi
path from Digital_Piracyi,t-1 to TV_Viewershipi,t, and the indirect path that
cant in the second half, thereby diluting the negative effect of digital
mediates Buzzi,t), the traditional ML will produce biased parameter es
piracy, and even creating an overall positive piracy effect for TV ratings.
timates, whereas the REML estimator can generate unbiased estimates
One interesting finding is that the direct effect is consistently sig
[15,23].
nificant for the first half, but insignificant for the second half. The result
suggests that US residents may engage in piracy for early episodes. After
3.3. Results consuming the early episodes, however, they either stick to the legiti
mate channel for later episodes or stop watching the drama, so that most
Tables 6(a), (b), (c), and (d) illustrate the analysis results of the of the remaining piracy users are non-US residents, who generate only
Google and YouTube search trends. Each table contains standardized attention effects for US residents. A similar consumption pattern was
coefficients for Eq. (1), Eq. (2), and Eq. (3) estimated by REML. found for Netflix content, where Netflix users tend to watch the original
6
Table 7
Analysis results with social media activities.
(a) Relationships among Viewers, Tweets, and (b) Relationships among TV_rating, Tweets, and (c) Relationships among Viewers, Engagements,
Piracy Piracy and Piracy
DV: Viewers DV: TV_rating DV: Viewers

Piracy − 0.021 (0.010)* Piracy − 0.010 (0.015) Piracy − 0.021 (0.010)*
Controls Yes Controls Yes Controls Yes
DV: Tweets DV: Tweets DV: Engagements
Piracy 0.071 (0.013)*** Piracy 0.071 (0.013)*** Piracy 0.081 (0.016)***
Tweets 0.083 (0.022)*** Tweets 0.128 (0.034)*** Engagements 0.054 (0.018)**
Piracy − 0.027 (0.010)** Piracy − 0.021 (0.016) Piracy − 0.026 (0.010)*
Indirect effect 0.006 Indirect 0.009 Indirect effect 0.004
effect
Direct effect − 0.027 Direct effect − 0.021 Direct effect − 0.026
Total effect − 0.021 Total effect − 0.012 Total effect − 0.021
Observations 1113 Observations 1113 Observations 1113
(d) Relationships among TV_rating, Engagements, (e) Relationships among Viewers, Interactions, and (f) Relationships among TV_rating, Interactions,
and Piracy Piracy and Piracy
DV: TV_rating DV: Viewers DV: TV_rating
Piracy − 0.010 (0.015) Piracy − 0.021 (0.010)* Piracy − 0.010 (0.015)
DV: Engagements DV: DV: Interactions
Interactions
Engagements 0.081 (0.028)** Interactions 0.053 (0.017)** Interactions 0.081 (0.027)**
Piracy − 0.018 (0.016) Piracy − 0.026 (0.010)** Piracy − 0.019 (0.016)
Indirect effect 0.007 Indirect effect 0.005 Indirect effect 0.007
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control variables include the IMDb star rating and runtime of each episode, and dummies
indicating the dramas, day of the week, broadcasting day, broadcasting week, broadcaster, and award.
airing after consuming early episodes on Netflix [42]. remaining as seeders such that they stop sharing the file after down
Prior empirical studies suggest that promotional effects of digital loading the whole packet. They may do so out of compunction to stop
piracy could be more substantial for movies in need of public awareness uploading, but may continue to engage in downloading. As such, seeders
[21,38]. Similarly, we examine the moderation effect of drama popu and leechers are heterogeneous in traits, and may have different impacts
larity (Tables 12–15). Specifically, we divide the entire sample into two on Internet buzz and TV viewership.
groups (high viewership vs. low viewership) with the cut-off viewer rate We replicate our main analysis for Seeds and Leeches. The results for
of 0.8 and replicate our analysis for each group.4 The positive effect of Seeds are consistent with our main results (Tables A1 and A2). The
Internet buzz is not significant both in the first and second halves of high negative direct effect of Seeds on TV viewership is neutralized by the
viewership dramas, suggesting the popularity offsets the word-of-mouth positive indirect effect that mediates Internet buzz. The indirect effect is
effect. However, in the case of low viewership dramas, the positive effect stronger for late episodes (Tables A3, A4, A5, and A6). Regarding
of Internet buzz is significant. Particularly, we find a stronger promo Leeches, however, only the indirect effect is significant (Tables A7 and
tional effect for their second half. A8). The results for early and late episodes are consistent with our main
We also compare the model estimates between procedural dramas results: the negative direct effect of Leeches on TV viewership is domi
(that do not need prior viewing of any other episode to understand) and nant for early episodes, but the indirect effect is dominant for late epi
serial dramas (that require viewing previous episodes) (Tables 16 and sodes (Tables A9, A10, A11, and A12). Assuming that followers of a
17). In the case of serial dramas, the story is interwoven across episodes; particular drama want others to watch it also, we may expect more
thus, Internet buzz about past episodes may carry a highly synergetic followers among seeders than leechers. Therefore, compared to seeders,
value for the consumption of the current episode. When we replicate our leechers are less likely to choose a legal channel if the piracy channel
analysis for a split sample, we find a significant word-of-mouth effect were not available, minimizing the substitution effect.
only for serial dramas.
Lastly, we examine the separate piracy effects of seeders and leechers
3.5. Robustness check
on TV viewership. Although leechers eventually become seeders once
they download the whole packet, there exist piracy users who avoid
3.5.1. Instrumental variable estimation
We supplement our main analysis by instrumenting Digital_Piracy to
alleviate endogeneity concerns due to possible reverse causality from
4
The viewer rate of all episodes of high (low) viewership dramas is higher Buzzi,t to Digital_Piracyi,t. We use the following instrumental variables for
(lower) than 0.8. Digital_Piracyi,t: (1) the number of available public BitTorrent trackers at
7
Table 8
Analysis results with search trends (First Half)**.
(a) Relationships among Viewers, (b) Relationships among TV_rating,
DV: Viewers DV: TV_rating

Piracy − 0.070 (0.014)*** Piracy − 0.094 (0.023)***
DV: Google DV: Google
Piracy 0.042 (0.019)* Piracy 0.042 (0.019)*
Google − 0.002 (0.031) Google 0.048 (0.051)
Piracy − 0.070 (0.014)*** Piracy − 0.096 (0.023)***
Indirect effect 0.000 Indirect effect 0.002
Direct effect − 0.070 Direct effect − 0.096
(c) Relationships among Viewers, (d) Relationships among TV_rating,

YouTube, and Piracy YouTube, and Piracy
Piracy − 0.070 (0.014)*** Piracy − 0.094 (0.023)***
DV: YouTube DV: YouTube
Piracy 0.058 (0.025)* Piracy 0.058 (0.025)*
YouTube 0.015 (0.023) YouTube 0.087 (0.038)*
Piracy − 0.071 (0.014)*** Piracy − 0.099 (0.023)***
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
dummies indicating the dramas, day of the week, broadcasting day, broad
casting week, broadcaster, and award.
from dramas. Given that ’Moviest,’ ’Pornt,’ ’Softwaret,’ and ’Gamest’

apparently do not have a direct influence on Buzzi,t (or vice versa), they
are eligible as instrumental variables for Digital_Piracyi,t.
The descriptive statistics for our instrumental variables and the
analysis results with the instrumental variables are provided in
Tables 18A, 18B, 19, and 20. The results are consistent with our main
findings.
Fig. 2. Results with social media activities. 3.5.2. Estimation with dimension reduction
Our analysis runs the model with each variable indicating Internet
time t (the variable is named ’Trackert’), and (2) the number of torrent Buzz (Google, YouTube, Tweets, Engagements, and Interactions) and finds
users who pirated movies, porn, software, and games in Brazil5 at time t the result consistency. As a robustness check, we apply the principal
(the variables are named ’Moviest,’ ’Pornt,’ ’Softwaret,’ and ’Gamest,’ component analysis (PCA) to reduce the dimension and replicate our
respectively). analysis with just one variable, Buzz. The result is consistent with our
A BitTorrent tracker is a special type of server, which assists in the main findings (Tables A13(a) and (b)). We also apply the PCA to inte
communication between peers on BitTorrent clients. Piracy users can grate TV_rating and Viewers to a single index, Viewership, for our analysis,
add such trackers to their clients, which increases download speeds. which generates consistent results with the main findings (Table A13
Given that the tracker is specific to BitTorrent, Trackert would affect (c)). Lastly, we obtain consistent results when using the integrated pi
Buzzi,t only through Digital_Piracyi,t. racy measure, Piracy_new, based on the PCA (Table A13(d)).
’Moviest,’ ’Pornt,’ ’Softwaret,’ and ’Gamest’ would either be positively
or negatively related to Digital_Piracyi,t. The larger base of piracy users in 3.5.3. Estimation with independent sample
other categories would encourage Digital_Piracyi,t, considering its ex We check the robustness of our empirical findings by replicating our
ternalities such as peer influence [17,48]. At the same time, the piracy analysis with an independent sample from South Korea. Similar to our
activities of other categories may absorb the interests of piracy users main sample, we obtained the viewership data of 26 TV dramas
5
Brazil is one of the largest countries with respect to torrent usage (https://
www.similarweb.com).
8
Table 9
Analysis results with social media activities (First Half)**.
(a) Relationships among Viewers, Tweets, and Piracy (b) Relationships among TV_rating, Tweets, and (c) Relationships among Viewers, Engagements,
Piracy and Piracy

Piracy − 0.070 (0.014)*** Piracy − 0.094 (0.023)*** Piracy − 0.070 (0.014)***
Tweets 0.068 (0.043) Tweets 0.122 (0.073)† Engagements 0.075 (0.046)
(d) Relationships among TV_rating, Engagements, and (e) Relationships among Viewers, Interactions, and (f) Relationships among TV_rating, Interactions,
Piracy Piracy and Piracy
DV: Engagements DV: Interactions DV: Interactions
Engagements 0.168 (0.078)* Interactions 0.069 (0.042) Interactions 0.153 (0.072)*
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table 10
Analysis results with search trends (Second Half).
(a) Relationships among Viewers, Google, and Piracy (b) Relationships among TV_rating, Google, and Piracy

Piracy 0.006 (0.011) Piracy 0.019 (0.017)
Piracy 0.069 (0.026)** Piracy 0.069 (0.026)**
Google 0.066 (0.019)*** Google 0.110 (0.029)***
Piracy 0.001 (0.011) Piracy 0.011 (0.017)
Direct effect 0.001 Direct effect 0.011
Total effect 0.006 Total effect 0.019
(c) Relationships among Viewers, YouTube, and Piracy (d) Relationships among TV_rating, YouTube, and Piracy
Piracy 0.006 (0.011) Piracy 0.019 (0.017)
Piracy 0.072 (0.027)** Piracy 0.072 (0.027)**
YouTube 0.022 (0.019) YouTube 0.061 (0.028)*
Piracy 0.004 (0.011) Piracy 0.014 (0.017)
(continued on next page)
9
Table 10 (continued )
(a) Relationships among Viewers, Google, and Piracy (b) Relationships among TV_rating, Google, and Piracy

Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table 11
Analysis results with social media activities (Second Half)*.
(a) Relationships among Viewers, Tweets, and Piracy (b) Relationships among TV_rating, Tweets, and (c) Relationships among Viewers, Engagements, and
Piracy Piracy

Piracy 0.006 (0.011) Piracy 0.019 (0.017) Piracy 0.006 (0.011)
Piracy 0.068 (0.022)** Piracy 0.068 (0.022)** Piracy 0.085 (0.027)***
Tweets 0.032 (0.022) Tweets 0.052 (0.033) Engagements 0.017 (0.017)
Direct effect 0.003 Direct effect 0.014 Direct effect 0.004
Total effect 0.005 Total effect 0.018 Total effect 0.005
(d) Relationships among TV_rating, Engagements, (e) Relationships among Viewers, Interactions, (f) Relationships among TV_rating, Interactions,
and Piracy and Piracy and Piracy
Engagements 0.025 (0.026) Interactions 0.017 (0.016) Interactions 0.026 (0.025)
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
broadcast in South Korea from June 2019 to August 2019 (Table A14),6 4. Discussion and conclusion
and scraped µTorrent throughout their broadcasting periods twice a day
at 12AM and 12PM (GMT+9). Table A15 illustrates the descriptive In this study, we empirically investigate the effects of TV drama pi
statistics of our variables used in this robustness check, and Table A16 racy on its legitimate viewership, which was silent in the research
provides the correlations between these variables. TV_rating and Seeds stream of digital piracy effects. The availability of the piracy channel
(or Leeches) are negatively correlated, but the relationship between might seduce potential consumers who would consume the contents of
Google and TV_rating, and the relationship between Google and Seeds (or the legitimate channel if the piracy channel were not available. How
Leeches) are all positive. Consistent with our main results, we find sig ever, such substitution effect could be counteracted by the attention
nificant mediation effects involving Google, which render the total pi effect, where the Internet buzz promoted by piracy users (who would not
racy effects insignificant (Table A17). consume the contents if the piracy channel were not available) may
draw the attention of potential consumers and may lead them to legal
consumption. Particularly, TV dramas are different from music or film in
that those who consume early episodes may generate Internet buzz for
later episodes. Furthermore, consumers who watched early episodes
6
would gain greater utility from watching late episodes than those who
Twenty-six Korean dramas were broadcast during our observation window,
did not, leading them to the legitimate channel for their on-time con
and we use them all for our analysis. The estimated number of viewers is not
available for the Korean data.
sumption. As such, enhanced access to early episodes afforded by piracy
10
Table 12 Table 13 (continued )

Analysis results with search trends (High Popularity & First Half)*,**. (a) Relationships among Viewers, (b) Relationships among TV_rating,
(a) Relationships among Viewers, (b) Relationships among TV_rating, Google, and Piracy Google, and Piracy
DV: Viewers DV: TV_rating Indirect effect 0.000 Indirect effect − 0.001
Piracy − 0.125 (0.020)*** Piracy − 0.171 (0.031)*** Direct effect 0.122 Direct effect 0.163
Controls Yes Controls Yes Total effect 0.122 Total effect 0.162
DV: Google DV: Google Observations 213 Observations 213
Piracy − 0.006 (0.025) Piracy − 0.006 (0.025)
Note: †p < 0.1.
Controls Yes Controls Yes *
DV: Viewers DV: TV_rating p < 0.05.
**
Google − 0.084 (0.045)† Google − 0.085 (0.070) p < 0.01.
***
Piracy − 0.125 (0.020)*** Piracy − 0.172 (0.031)*** p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control
Controls Yes Controls Yes variables include the IMDb star rating and runtime of each episode, and
Indirect effect 0.001 Indirect effect 0.000 dummies indicating the dramas, day of the week, broadcasting day, broad
Direct effect − 0.125 Direct effect − 0.172 casting week, broadcaster, and award.
(c) Relationships among Viewers, (d) Relationships among TV_rating, Table 14

YouTube, and Piracy YouTube, and Piracy Analysis results with search trends (Low Popularity & First Half)*.
Piracy − 0.125 (0.020)*** Piracy − 0.171 (0.031)*** (a) Relationships among Viewers, (b) Relationships among TV_rating,
Controls Yes Controls Yes Google, and Piracy Google, and Piracy
Piracy − 0.002 (0.015) Piracy − 0.002 (0.015)
Piracy 0.006 (0.010) Piracy 0.011 (0.022)
YouTube − 0.101 (0.075) YouTube − 0.155 (0.117)
Piracy 0.099 (0.028)*** Piracy 0.099 (0.028)***
Piracy − 0.125 (0.020)*** Piracy − 0.172 (0.031)***
Google − 0.018 (0.014) Google 0.065 (0.042)
Piracy 0.009 (0.010) Piracy 0.006 (0.023)
Indirect effect − 0.002 Indirect effect 0.006
Note: †p < 0.1. Direct effect 0.009 Direct effect 0.006
* Total effect 0.007 Total effect 0.012
p < 0.05.
** Observations 249 Observations 249
p < 0.01.
***
p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control (c) Relationships among Viewers, (d) Relationships among TV_rating,
variables include the IMDb star rating and runtime of each episode, and YouTube, and Piracy YouTube, and Piracy
dummies indicating the dramas, day of the week, broadcasting day, broad DV: Viewers DV: TV_rating
casting week, broadcaster, and award. Piracy 0.006 (0.010) Piracy 0.011 (0.022)
Piracy 0.158 (0.054)** Piracy 0.158 (0.054)**
Table 13 Controls Yes Controls Yes
Analysis results with search trends (High Popularity & Second Half)*, **. DV: Viewers DV: TV_rating
YouTube − 0.009 (0.010) YouTube 0.093 (0.022)***
(a) Relationships among Viewers, (b) Relationships among TV_rating,
Piracy 0.008 (0.010) Piracy − 0.001 (0.022)
DV: Viewers DV: TV_rating Indirect effect − 0.001 Indirect effect 0.015
Piracy 0.122 (0.036)*** Piracy 0.162 (0.046)*** Direct effect 0.008 Direct effect − 0.001
Controls Yes Controls Yes Total effect 0.006 Total effect 0.014
Piracy 0.031 (0.072) Piracy 0.031 (0.072)
Note: †p < 0.1.
*
**
Google 0.065 (0.036)† Google 0.086 (0.046)† p < 0.01.
***
Piracy 0.121 (0.036)*** Piracy 0.160 (0.046)*** p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control
Controls Yes Controls Yes variables include the IMDb star rating and runtime of each episode, and
Indirect effect 0.002 Indirect effect 0.003 dummies indicating the dramas, day of the week, broadcasting day, broad
Direct effect 0.121 Direct effect 0.160 casting week, broadcaster, and award.
channels could help the legitimate consumption of late episodes.7 As a
(c) Relationships among Viewers, (d) Relationships among TV_rating, result, we find insignificant overall piracy effects for both US and Korean
YouTube, and Piracy YouTube, and Piracy TV dramas.
Piracy 0.122 (0.036)*** Piracy 0.162 (0.046)***
Our study contributes to the research stream of digital piracy effects.
Controls Yes Controls Yes Prior empirical studies in Information Systems reveal that movie or
DV: YouTube DV: YouTube music piracy may or may not harm the corresponding content industry.
Piracy − 0.011 (0.027) Piracy − 0.011 (0.027)
YouTube − 0.028 (0.098) YouTube 0.120 (0.124) 7
Such mechanism can be found in other digital goods when previous con
Piracy 0.122 (0.036)*** Piracy 0.163 (0.046)***
sumption drives later legitimate consumption of the same goods. For example,
listening to a pirated song can make people buy the album.
11
Table 15 Table 16 (continued )

Analysis results with search trends (Low Popularity & Second Half). (a) Relationships among Viewers, (b) Relationships among TV_rating,
(a) Relationships among Viewers, (b) Relationships among TV_rating, Google, and Piracy Google, and Piracy
Piracy − 0.011 (0.024) Piracy 0.039 (0.049)
DV: Viewers DV: TV_rating Controls Yes Controls Yes
Piracy − 0.007 (0.006) Piracy 0.007 (0.012) Indirect effect 0.000 Indirect effect − 0.002
Controls Yes Controls Yes Direct effect − 0.011 Direct effect 0.039
DV: Google DV: Google Total effect − 0.012 Total effect 0.036
Piracy 0.085 (0.026)*** Piracy 0.085 (0.026)*** Observations 357 Observations 357
Note: †p < 0.1.
DV: Viewers DV: TV_rating *
Google 0.016 (0.015) Google 0.043 (0.027) p < 0.05.
**
Piracy − 0.008 (0.007) Piracy 0.003 (0.012) p < 0.01.
***
Controls Yes Controls Yes p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control
Indirect effect 0.001 Indirect effect 0.004 variables include the IMDb star rating and runtime of each episode, and
Direct effect − 0.008 Direct effect 0.003 dummies indicating the dramas, day of the week, broadcasting day, broad
Total effect − 0.007 Total effect 0.007 casting week, broadcaster, and award.

YouTube, and Piracy YouTube, and Piracy Table 17
DV: Viewers DV: TV_rating Analysis results with search trends (Serial Dramas).
Piracy − 0.007 (0.006) Piracy 0.007 (0.012)
Controls Yes Controls Yes (a) Relationships among Viewers, (b) Relationships among TV_rating,
DV: YouTube DV: YouTube Google, and Piracy Google, and Piracy
Piracy 0.097 (0.037)** Piracy 0.097 (0.037)**
Piracy − 0.025 (0.010)* Piracy − 0.018 (0.015)
YouTube 0.013 (0.010) YouTube 0.038 (0.019)*
Piracy − 0.008 (0.006) Piracy 0.003 (0.012)
Piracy 0.052 (0.021)* Piracy 0.052 (0.021)*
Direct effect − 0.008 Direct effect 0.003
Google 0.097 (0.018)*** Google 0.161 (0.024)***
Total effect − 0.007 Total effect 0.007
Piracy − 0.030 (0.010)** Piracy − 0.026 (0.014)†
Note: †p < 0.1. Indirect effect 0.005 Indirect effect 0.008
* Direct effect − 0.030 Direct effect − 0.026
p < 0.05.
** Total effect − 0.025 Total effect − 0.018
p < 0.01.
*** Observations 756 Observations 756
variables include the IMDb star rating and runtime of each episode, and (c) Relationships among Viewers, (d) Relationships among TV_rating,
dummies indicating the dramas, day of the week, broadcasting day, broad YouTube, and Piracy YouTube, and Piracy
casting week, broadcaster, and award. DV: Viewers DV: TV_rating
Piracy − 0.025 (0.010)* Piracy − 0.018 (0.015)
Piracy 0.058 (0.023)* Piracy 0.058 (0.023)*
Table 16
Analysis results with search trends (Procedural Dramas)*,**,**. DV: Viewers DV: TV_rating
(a) Relationships among Viewers, (b) Relationships among TV_rating, YouTube 0.034 (0.016)* YouTube 0.089 (0.023)***
Google, and Piracy Google, and Piracy Piracy − 0.027 (0.010)** Piracy − 0.023 (0.014)
DV: Viewers DV: TV_rating Indirect effect 0.002 Indirect effect 0.005
Piracy − 0.012 (0.024) Piracy 0.036 (0.049) Direct effect − 0.027 Direct effect − 0.023
Controls Yes Controls Yes Total effect − 0.025 Total effect − 0.018
Piracy 0.008 (0.016) Piracy 0.008 (0.016)
Controls Yes Controls Yes Note: †p < 0.1.
*
Google 0.028 (0.082) Google 0.070 (0.170) **
p < 0.01.
Piracy − 0.012 (0.024) Piracy 0.036 (0.049) ***
dummies indicating the dramas, day of the week, broadcasting day, broad
Total effect − 0.012 Total effect 0.036 casting week, broadcaster, and award.

YouTube, and Piracy YouTube, and Piracy
This study is one of the first that empirically investigates the effect of TV
DV: Viewers DV: TV_rating drama piracy. Unlike movies or music, TV drama is a series of episodes,
Piracy − 0.012 (0.024) Piracy 0.036 (0.049) where piracy of early episodes may promote the legal consumption of
Controls Yes Controls Yes later episodes. We show that such effects are stronger for later episodes,
as the Internet buzz snowballs over time. Further, such effects are
Piracy − 0.009 (0.007) Piracy − 0.009 (0.007)
Controls Yes Controls Yes stronger for less popular and serial dramas.
DV: Viewers DV: TV_rating Comparing procedural dramas and serial dramas may cast an inter
YouTube 0.015 (0.175) YouTube 0.253 (0.361) esting research question for other piracy contexts, such as movie piracy.
Recent empirical studies regarding movie piracy identify conditional
12
Table 18A
. Descriptive statistics for instrumental variables.
Variable Mean Std. Dev. Min Max
Tracker (log) 2.175 1.114 0 4.444

Movies (log) 14.280 0.106 13.923 14.531
Porn (log) 10.976 0.107 10.533 11.162
Software (log) 11.307 0.107 10.799 11.495
Games (log) 11.906 0.104 11.413 12.066
Table 18B
Correlations for instrumental variables.
(1) (2) (3) (4) (5)
Tracker (log) 1
Movies (log) 0.034 1
Porn (log) 0.008 0.778 1
Software (log) 0.001 0.557 0.661 1
Games (log) − 0.042 0.613 0.654 0.786 1
Table 19 but in an offline setting [39]. They found that the sales of counterfeit
Analysis results with instrumental variables (Search Trends). shoes created larger sales of authentic shoes. This is because those
(a) Relationships among Viewers, (b) Relationships among TV_rating, counterfeits can increase brand awareness for authentic products with
Google, and PiracyIV Google, and PiracyIV mass advertising. Given that there are consumers who do not want to or
DV: Viewers DV: TV_rating have no idea of how to purchase counterfeits, consumers who bought
PiracyIV − 0.064 (0.014)*** PiracyIV − 0.065 (0.022)** counterfeits may play a role in advertising for those who stick to pur
Controls Yes Controls Yes chase authentic products. Our findings are consistent with their argu
DV: Google DV: Google ment in that there are consumers who only use legitimate channels to
PiracyIV 0.053 (0.023)* PiracyIV 0.053 (0.023)*
watch TV dramas. These consumers may be exposed to Internet buzz
DV: Viewers DV: TV_rating regarding the dramas, which will entice them to watch such dramas.
Google 0.106 (0.018)*** Google 0.170 (0.028)*** From a practical point of view, the negative effect of digital piracy
PiracyIV − 0.070 (0.014)*** PiracyIV − 0.074 (0.021)** might be limited, particularly for developed countries such as the US.
Under strict anti-Internet piracy laws, US residents are not allowed to
Direct effect − 0.070 Direct effect − 0.074 consume pirated content on BitTorrent clients. Statistics show that most
Total effect − 0.064 Total effect − 0.065 digital piracy activities occur outside the US, which attests to the pos
Observations 1113 Observations 1113 sibility of a limited substitution effect. Those piracy users outside the US,
however, could contribute to the Internet buzz of TV dramas, which may
YouTube, and PiracyIV YouTube, and PiracyIV
entice the attention of US residents. The indirect effect could produce an
DV: Viewers DV: TV_rating overall positive effect of digital piracy, such as the second-half episodes
PiracyIV − 0.064 (0.014)*** PiracyIV − 0.065 (0.022)** of the TV dramas in our analysis. Further, copyright holders may release
Controls Yes Controls Yes previous episodes for free if their serial drama suffers from eliciting
public attention. By doing so, they could exploit the attention effects for
PiracyIV 0.051 (0.026)* PiracyIV 0.051 (0.026)*
Controls Yes Controls Yes their later episodes.
DV: Viewers DV: TV_rating We conclude by articulating a limitation and providing suggestions
YouTube 0.036 (0.016)* YouTube 0.090 (0.026)** for future research. Although we focus on BitTorrent to investigate pi
PiracyIV − 0.066 (0.014)*** PiracyIV − 0.069 (0.022)** racy effects, there are other piracy channels. As such, our finding does
not reflect all piracy effects. Furthermore, there are other revenue
Direct effect − 0.066 Direct effect − 0.069 sources for broadcasters, such as video-on-demand services or paid
Total effect − 0.064 Total effect − 0.065 subscriptions, for replays on their websites. The piracy effects on those
Observations 1113 Observations 1113 services could be another meaningful topic for the media industry. Also,
Note: †p < 0.1. we may underestimate the potential benefit of TV drama piracy because
*
p < 0.05. we do not consider other positive mediation effects driven by piracy,
**
p < 0.01. such as offline buzz. Lastly, the attention effect could depend on cultural
***
p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control orientation, where collectivism might enforce the effect, considering
variables include the IMDb star rating and runtime of each episode, and that people from collectivistic cultures are more likely to follow others’
dummies indicating the dramas, day of the week, broadcasting day, broad opinions. Further study may be conducted to explore the role of cultural
casting week, broadcaster, and award. orientation in the piracy effect.8
promotional effects of pirated content. Along with this research stream,

our finding may indicate that, in the case of movie series, piracy activ
ities of a prior version may help the success of a newer version when
8
their stories are interwoven. Our data do not allow us to investigate such cultural effects, as the U.S. and
Korean data are collected in different time windows. Time dummies in our
A paper in the Marketing discipline introduces a similar discussion
model capture all cultural differences.
13
Table 20
Analysis results with instrumental variables (Social Media Activities).
(a) Relationships among Viewers, Tweets, and (b) Relationships among TV_rating, Tweets, and (c) Relationships among Viewers, Engagements, and
PiracyIV PiracyIV PiracyIV

PiracyIV − 0.064 (0.014)*** PiracyIV − 0.065 (0.022)** PiracyIV − 0.064 (0.014)***
PiracyIV 0.041 (0.018)* PiracyIV 0.041 (0.018)* PiracyIV 0.040 (0.021)†
PiracyIV − 0.069 (0.014)*** PiracyIV − 0.073 (0.022)** PiracyIV − 0.067 (0.014)***
effect
and PiracyIV PiracyIV and PiracyIV
PiracyIV − 0.065 (0.022)** PiracyIV − 0.064 (0.014)*** PiracyIV − 0.065 (0.022)**
Interactions
PiracyIV 0.040 (0.021)† PiracyIV 0.043 (0.022)† PiracyIV 0.043 (0.022)†
PiracyIV − 0.070 (0.022)** PiracyIV − 0.068 (0.014)*** PiracyIV − 0.070 (0.022)**
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
CRediT authorship contribution statement Acknowledgments
Dongyeon Kim: Conceptualization, Data curation, Formal analysis, This work was supported by the Yonsei Signature Research Cluster
Methodology, Software. Kyuhong Park: Conceptualization, Data cura Program [grant number 2021-22-0006]; the Catholic University of
tion, Investigation, Software. Youngsok Bang: Conceptualization, Su Korea Research Fund of 2022; and the Inha University Research Fund of
pervision, Writing – original draft, Writing – review & editing. 2022.
Declarations of Competing Interest
None.
Appendix 1. Results of PVAR analysis
• Piracy, Viewer, and Google
Selection order criteria
lag CD J J p-value MBIC MAIC MQIC

1 0.999873 13.638 0.984 − 168.579 − 40.362 − 89.467
2 0.999911 10.741 0.905 − 110.736 − 25.259 − 57.996
3 0.999914 1.356 0.998 − 59.383 − 16.644 − 33.013
14
Select lag 1
DV: Viewers
Viewers (L1) 1.029861 (0.034246)***
Piracy (L1) 0.000002 (0.000005)
Google (L1) 0.000025 (0.000422)
DV: Piracy
Viewers (L1) − 49.69603 (272.5421)
Piracy (L1) 0.287449 (0.048878)***
Google (L1) 2.804663 (3.667775)
DV: Google
Viewers (L1) 2.44716 (7.443889)
Piracy (L1) 0.0031921 (0.000795)***
Google (L1) 0.115086 (0.142410)
Note: *p < 0.05, **p < 0.01, ***p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control variables include the IMDb star rating and
runtime of each episode, and dummies indicating the dramas, day of the week, broadcasting day, broadcasting week, broadcaster, and award.
Panel VAR-Granger causality Wald test
Equation
Variables Viewers Piracy Google
Viewers – 0.033 0.108
Piracy 0.104 – 16.122***
Google 0.004 0.585 –
Note: *p < 0.05, **p < 0.01, ***p < 0.001, Two-tailed tests; The value represents chi-square statistics.
• Piracy, TV_rating, and Google

1 0.9994876 17.73601 0.9113813 − 164.4805 − 36.26399 − 85.36936
2 0.9996207 7.666163 0.9832028 − 113.8115 − 28.33384 − 61.07075
3 0.9996091 1.877785 0.993256 − 58.86105 − 16.12222 − 32.49067
Select lag 1
DV: TV_rating
TV_rating (L1) 0.962133 (0.032830)***
Piracy (L1) 0.000002 (0.000001)
Google (L1) 0.000017 (0.000126)
DV: Piracy
TV_rating (L1) 448.4149 (943.6522)
Piracy (L1) 0.295478 (0.045746)***
Google (L1) 2.471372 (3.273831)
DV: Google
TV_rating (L1) 4.861389 (23.0241)
Piracy (L1) 0.002721 (0.000730)***
Google (L1) 0.135303 (0.122779)
Equation
Variables TV_rating Piracy Google
TV_rating – 0.226 0.045
Piracy 2.151 – 13.887***
Google 0.018 0.570 –
15
• Piracy, Viewer, and YouTube

1 0.999865 13.75474 0.983536 − 168.462 − 40.2453 − 89.3506
2 0.999886 10.25139 0.923391 − 111.226 − 25.7486 − 58.4855
3 0.999877 2.423578 0.982869 − 58.3153 − 15.5764 − 31.9449
Select lag 1
DV: Viewers
Viewers (L1) 1.023312 (0.027579)***
Piracy (L1) 0.000010 (0.000004)
YouTube (L1) 0.000113 (0.000181)
DV: Piracy
Viewers (L1) 212.6316 (221.6016)
Piracy (L1) 0.306387 (0.042472)***
YouTube (L1) 0.552309 (2.225186)
DV: YouTube
Viewers (L1) − 4.881886 (7.560149)
Piracy (L1) 0.006561 (0.001114)***
YouTube (L1) 0.176709 (0.091146)†
Equation
Variables Viewers Piracy YouTube
Viewers – 0.921 0.417
Piracy 0.050 – 34.715***
YouTube 0.391 0.062 –
• Piracy, TV_rating, and YouTube
Lag CD J J p-value MBIC MAIC MQIC

1 0.9993918 23.87835 0.6370551 − 158.3382 − 30.12165 − 79.22702
2 0.9995016 11.9771 0.8484174 − 109.5006 − 24.0229 − 56.75981
3 0.9994226 4.714042 0.8584897 − 56.02479 − 13.28596 − 29.65441
Select lag 1
DV: TV_rating
TV_rating (L1) 0.973552 (0.030062)***
Piracy (L1) 0.000001 (0.000001)
YouTube (L1) 0.000024 (0.000063)
DV: Piracy
TV_rating (L1) 746.172 (836.7969)
Piracy (L1) 0.317372 (0.043137)***
YouTube (L1) 0.581888 (2.214889)
DV: YouTube
TV_rating (L1) − 34.82867 (32.02116)
Piracy (L1) 0.006911 (0.001045)***
YouTube (L1) 0.207783 (0.087923)*
16
Equation
Variables TV_rating Piracy YouTube
TV_rating – 0.795 1.183
Piracy 0.925 – 43.759***
YouTube 0.144 0.069 –
• Piracy, Viewer, and Tweets

1 0.998647 35.05002 0.137603 − 145.603 − 18.95 − 67.5899
2 0.998839 17.0832 0.517392 − 103.352 − 18.9168 − 51.3434
3 0.99527 4.270381 0.892733 − 55.9472 − 13.7296 − 29.9429
Select lag 1
DV: Viewers
Viewers (L1) .9368295 (0.0810025)***
Piracy (L1) − 0.000006 (0.000008)
Tweets (L1) 0.000015 (0.000002)***
DV: Piracy
Viewers (L1) 409.7804 (254.0345)
Piracy (L1) 0.303472 (0.043922)***
Tweets (L1) 0.003117 (0.006342)
DV: Tweets
Viewers (L1) − 567.2154 (740.1133)
Piracy (L1) 0.582642 (0.113743)***
Tweets (L1) 0.015549 (0.037545)
Equation
Variables Viewers Piracy Tweets
Viewers – 2.602 0.587
Piracy 0.609 – 26.239***
Tweets 34.690*** 0.242 –
• Piracy, TV_rating, and Tweets
lag CD J J pvalue MBIC MAIC MQIC

1 0.995148 47.73102 0.008227 − 132.922 − 6.26899 − 54.9089
2 0.995506 16.70539 0.54344 − 103.73 − 19.2946 − 51.7212
3 0.938102 6.7162 0.66664 − 53.5014 − 11.2838 − 27.4971
Select lag 1
DV: TV_rating
TV_rating (L1) .9477863 (0.0510078)***
Piracy (L1) − 0.000001 (0.000001)
Tweets (L1) 0.000004 (0.000001)***
DV: Piracy
TV_rating (L1) 2060.68 (812.2736)*
Piracy (L1) 0.300493 (0.043767)***
Tweets (L1) 0.007530 (0.006181)
17
(continued )
DV: Tweets
TV_rating (L1) − 2209.616 (2485.885)
Piracy (L1) 0.572870 (0.110251)***
Tweets (L1) 0.007093 (0.041324)
Equation
Variables TV_rating Piracy Tweets
TV_rating – 6.436* 0.790
Piracy 0.266 – 26.999***
Tweets 79.238*** 1.484 –
• Piracy, Viewer, and Engagements

1 0.998666 31.16829 0.264285 − 149.485 − 22.8317 − 71.4717
2 0.998831 14.83431 0.673311 − 105.601 − 21.1657 − 53.5923
3 0.998816 7.248646 0.611249 − 52.9689 − 10.7514 − 26.9647
Select lag 1
DV: Viewers
Viewers (L1) 0.957026 (0.071051)***
Piracy (L1) − 0.000007 (0.000007)
Engagements (L1) 0.001722 (0.000283)***
DV: Piracy
Viewers (L1) 337.6706 (273.8305)
Piracy (L1) 0.304447 (0.044168)***
Engagements (L1) 0.178616 (0.809716)
DV: Engagements
Viewers (L1) 0.680687 (10.4661)
Piracy (L1) 0.004525 (0.001573)**
Engagements (L1) 0.079093 (0.086362)
Equation
Variables Viewers Piracy Engagements
Viewers – 1.521 0.004
Piracy 0.958 – 8.280**
Engagements 37.002*** 0.049 –
• Piracy, TV_rating, and Engagements

1 0.994998 46.28619 0.01185 − 134.367 − 7.71381 − 56.3538
2 0.995598 23.04326 0.188933 − 97.3919 − 12.9567 − 45.3834
3 0.995815 3.344697 0.949049 − 56.8729 − 14.6553 − 30.8686
18
Select lag 1
DV: TV_rating
TV_rating (L1) 0.926211 (0.044653)***
Piracy (L1) − 0.000008 (0.000001)
Engagements (L1) 0.000492 (0.000057)***
DV: Piracy
TV_rating (L1) 1802.089 (890.7209)
Piracy (L1) 0.307792 (0.044037)***
Engagements (L1) 0.648465 (0.867999)
DV: Engagements
TV_rating (L1) − 21.24553 (37.8199)
Piracy (L1) 0.004382 (0.001628)***
Engagements (L1) 0.046139 (0.083579)
Equation
Variables TV_rating Piracy Engagements
TV_rating – 4.093* 0.316
Piracy 0.286 – 7.246**
Engagements 73.558*** 0.558 –
• Piracy, Viewer, and Interactions
Lag CD J J p-value MBIC MAIC MQIC

1 0.998683 32.00208 0.232007 − 148.651 − 21.9979 − 70.6379
2 0.99885 14.52375 0.694366 − 105.911 − 21.4763 − 53.9029
3 0.998196 5.786911 0.761041 − 54.4307 − 12.2131 − 28.4264
Select lag 1
DV: Viewers
Viewers (L1) 0.955043 (0.072486)***
Piracy (L1) − 0.000007 (0.000007)
Interactions (L1) 0.001565 (0.000275)***
DV: Piracy
Viewers (L1) 349.9157 (272.2764)
Piracy (L1) 0.303513 (0.044145)***
Interactions (L1) 0.161979 (0.721907)
DV: Interactions
Viewers (L1) 0.005742 (11.22428)
Piracy (L1) 0.005091 (0.001674)**
Interactions (L1) 0.067907 (0.076737)
Equation
Variables Viewers Piracy Interactions
Viewers – 1.652 0.001
Piracy 0.923 – 9.247**
Interactions 32.284*** 0.050 –
19
• Piracy, TV_rating, and Interactions

1 0.99503 46.39626 0.011529 − 134.257 − 7.60374 − 56.2437
2 0.995615 23.12976 0.185653 − 97.3054 − 12.8702 − 45.2969
3 0.995633 3.566458 0.937566 − 56.6511 − 14.4335 − 30.6469
Select lag 1
DV: TV_rating
TV_rating (L1) 0.927027 (0.045714)***
Piracy (L1) − 0.000008 (0.000001)
Interactions (L1) 0.000448 (0.000052)
DV: Piracy
TV_rating (L1) 1820.555 (885.0441)*
Piracy (L1) 0.306885 (0.043973)***
Interactions (L1) 0.578269 (0.762740)
DV: Interactions
TV_rating (L1) − 24.92328 (40.55082)
Piracy (L1) 0.004904 (0.001735)**
Interactions (L1) 0.038601 (0.076648)
Equation
Variables TV_rating Piracy Interactions
TV_rating – 4.231* 0.378
Piracy 0.302 – 7.988**
Interactions 74.184*** 0.575 –
Appendix 2. Analysis results of Seeds and Leeches
Table A1
Analysis results with search trends (Seeds).
(a) Relationships among Viewers, Google, and Seeds (b) Relationships among TV_rating, Google, and Seeds

Seeds − 0.026 (0.008)** Seeds − 0.012 (0.013)
Seeds 0.052 (0.014)*** Seeds 0.052 (0.014)***
Google 0.107 (0.018)*** Google 0.168 (0.028)***
Seeds − 0.032 (0.008)*** Seeds − 0.021 (0.013)
(c) Relationships among Viewers, YouTube, and Seeds (d) Relationships among TV_rating, YouTube, and Seeds
Seeds − 0.026 (0.008)** Seeds − 0.012 (0.013)
Seeds 0.056 (0.016)*** Seeds 0.056 (0.016)***
YouTube 0.037 (0.017)* YouTube 0.088 (0.026)***
Seeds − 0.028 (0.008)*** Seeds − 0.017 (0.013)
20
Table A1 (continued )

Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A2
Analysis results with social media activities (Seeds)*.
(a) Relationships among Viewers, Tweets, and Seeds (b) Relationships among TV_rating, Tweets, and (c) Relationships among Viewers, Engagements,
Seeds and Seeds

Seeds − 0.026 (0.008)** Seeds − 0.012 (0.013) Seeds − 0.026 (0.008)**
Seeds 0.047 (0.011)*** Seeds 0.047 (0.011)*** Seeds 0.056 (0.014)***
Seeds − 0.031 (0.008)*** Seeds − 0.019 (0.013) Seeds − 0.030 (0.009)***
effect
and Seeds Seeds and Seeds
Seeds − 0.012 (0.013) Seeds − 0.026 (0.008)** Seeds − 0.012 (0.013)
Interactions
Seeds − 0.018 (0.013) Seeds − 0.030 (0.009)*** Seeds − 0.018 (0.013)
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A3
Analysis results with search trends (Seeds, First Half).

Seeds − 0.059 (0.011)*** Seeds − 0.070 (0.019)***
Seeds 0.042 (0.015)** Seeds 0.042 (0.015)**
Google 0.002 (0.031) Google 0.051 (0.052)
Seeds − 0.059 (0.011)*** Seeds − 0.072 (0.019)***
21

Seeds − 0.059 (0.011)*** Seeds − 0.070 (0.019)***
Seeds 0.062 (0.021)** Seeds 0.062 (0.021)**
YouTube 0.019 (0.023) YouTube 0.090 (0.039)*
Seeds − 0.060 (0.011)*** Seeds − 0.076 (0.019)***
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A4
Analysis results with social media activities (Seeds, First Half)*,**.
(a) Relationships among Viewers, Tweets, and Seeds (b) Relationships among TV_rating, Tweets, and (c) Relationships among Viewers, Engagements, and
Seeds Seeds

Seeds − 0.059 (0.011)*** Seeds − 0.070 (0.018)*** Seeds − 0.059 (0.011)***
and Seeds and Seeds and Seeds
Engagements 0.143 (0.077)† Interactions 0.058 (0.042) Interactions 0.128 (0.071)†
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
22
Table A5
Analysis results with search trends (Seeds, Second Half).

Seeds − 0.002 (0.010) Seeds 0.014 (0.015)
Seeds 0.073 (0.023)** Seeds 0.073 (0.023)**
Google 0.068 (0.020)** Google 0.111 (0.029)***
Seeds − 0.007 (0.010) Seeds 0.006 (0.015)
Seeds − 0.002 (0.010) Seeds 0.014 (0.015)
Seeds 0.075 (0.024)** Seeds 0.075 (0.024)**
YouTube 0.024 (0.019) YouTube 0.062 (0.028)*
Seeds − 0.004 (0.010) Seeds 0.010 (0.015)
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A6
Analysis results with social media activities (Seeds, Second Half)**.
(a) Relationships among Viewers, Tweets, and Seeds (b) Relationships among TV_rating, Tweets, and (c) Relationships among Viewers, Engagements, and
Seeds Seeds

Seeds − 0.002 (0.010) Seeds 0.014 (0.015) Seeds − 0.002 (0.010)
Seeds 0.045 (0.019)* Seeds 0.045 (0.019)* Seeds 0.045 (0.019)*
Seeds − 0.005 (0.010) Seeds 0.011 (0.015) Seeds − 0.004 (0.010)
Direct effect − 0.005 Direct effect 0.011 Direct effect − 0.004
Total effect − 0.003 Total effect 0.013 Total effect − 0.003
and Seeds and Seeds and Seeds
Seeds 0.014 (0.015) Seeds − 0.002 (0.010) Seeds 0.014 (0.015)
Seeds 0.045 (0.019)* Seeds 0.045 (0.019)* Seeds 0.045 (0.019)*
Seeds 0.012 (0.015) Seeds − 0.004 (0.010) Seeds 0.012 (0.015)
Direct effect 0.012 Direct effect − 0.004 Direct effect 0.012
Total effect 0.013 Total effect − 0.003 Total effect 0.013
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
23
Table A7
Analysis results with search trends (Leeches)*,**.
(a) Relationships among Viewers, Google, and Leeches (b) Relationships among TV_rating, Google, and Leeches

Leeches 0.004 (0.010) Leeches − 0.001 (0.016)
Leeches 0.022 (0.017) Leeches 0.022 (0.017)
Google 0.100 (0.018)*** Google 0.163 (0.028)***
Leeches 0.002 (0.010) Leeches − 0.005 (0.016)
Direct effect 0.002 Direct effect − 0.005
Total effect 0.004 Total effect − 0.001
(c) Relationships among Viewers, YouTube, and Leeches (d) Relationships among TV_rating, YouTube, and Leeches
Leeches 0.004 (0.010) Leeches − 0.001 (0.016)
Leeches 0.014 (0.019) Leeches 0.014 (0.019)
YouTube 0.031 (0.017)† YouTube 0.085 (0.026)***
Leeches 0.004 (0.010) Leeches − 0.002 (0.016)
Direct effect 0.004 Direct effect − 0.002
Total effect 0.004 Total effect − 0.001
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A8
Analysis results with social media activities (Leeches).
Leeches Leeches and Leeches

Leeches 0.004 (0.010) Leeches − 0.001 (0.016) Leeches 0.004 (0.010)
Leeches 0.091 (0.014)*** Leeches 0.091 (0.014)*** Leeches 0.100 (0.017)***
Tweets 0.074 (0.022)** Tweets 0.125 (0.035)*** Engagements 0.046 (0.018)*
Leeches − 0.003 (0.011) Leeches − 0.013 (0.017) Leeches − 0.001 (0.010)
effect
Total effect − 0.004 Total effect − 0.002 Total effect 0.004
and Leeches Leeches and Leeches
Leeches − 0.001 (0.016) Leeches 0.004 (0.010) Leeches − 0.001 (0.016)
Interactions
Leeches − 0.010 (0.016) Leeches − 0.001 (0.010) Leeches − 0.010 (0.016)
24

Total effect − 0.002 Total effect 0.004 Total effect − 0.002
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A9
Analysis results with search trends (Leeches, First Half)*.

Leeches − 0.049 (0.016)** Leeches − 0.092 (0.026)***
Leeches 0.005 (0.021) Leeches 0.005 (0.021)
Google − 0.016 (0.031) Google 0.030 (0.051)
Leeches − 0.049 (0.016)** Leeches − 0.093 (0.026)***
Leeches − 0.049 (0.016)** Leeches − 0.092 (0.026)***
Leeches − 0.003 (0.029) Leeches − 0.003 (0.029)
YouTube 0.003 (0.023) YouTube 0.070 (0.038)†
Leeches − 0.049 (0.016)** Leeches − 0.092 (0.026)***
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A10
Analysis results with social media activities (Leeches, First Half).
Leeches Leeches Leeches

Leeches − 0.049 (0.016)** Leeches − 0.092 (0.026)*** Leeches − 0.049 (0.016)**
Tweets 0.049 (0.044) Tweets 0.126 (0.074)† Engagements 0.055 (0.047)
Leeches − 0.055 (0.017)*** Leeches − 0.109 (0.028)*** Leeches − 0.055 (0.016)***
25

Leeches Leeches Leeches
and Leeches and Leeches and Leeches
Leeches − 0.092 (0.026)*** Leeches − 0.049 (0.016)** Leeches − 0.092 (0.026)***
Engagements 0.169 (0.079)* Interactions 0.051 (0.043) Interactions 0.154 (0.073)*
Leeches − 0.110 (0.027)*** Leeches − 0.055 (0.016)*** Leeches − 0.110 (0.027)***
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A11
Analysis results with search trends (Leeches, Second Half)**.

Leeches 0.022 (0.011)† Leeches 0.021 (0.017)
Leeches 0.034 (0.026) Leeches 0.034 (0.026)
Google 0.064 (0.019)*** Google 0.111 (0.029)***
Leeches 0.020 (0.011)† Leeches 0.018 (0.017)
Leeches 0.022 (0.011)† Leeches 0.021 (0.017)
Leeches 0.038 (0.028) Leeches 0.038 (0.028)
YouTube 0.021 (0.018) YouTube 0.062 (0.028)*
Leeches 0.021 (0.011)† Leeches 0.019 (0.017)
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
26
Table A12
Analysis results with social media activities (Leeches, Second Half)*,**.

Leeches 0.022 (0.011)† Leeches 0.021 (0.017) Leeches 0.022 (0.011)†
Leeches 0.019 (0.012)† Leeches 0.016 (0.018) Leeches 0.020 (0.012)†
(d) Relationships among TV_rating, Engagements, and (e) Relationships among Viewers, Interactions, and (f) Relationships among TV_rating, Interactions,
Leeches 0.021 (0.017) Leeches 0.022 (0.011)† Leeches 0.021 (0.017)
Leeches 0.018 (0.018) Leeches 0.020 (0.012)† Leeches 0.018 (0.018)
Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A13
Analysis results with integrated measures.
(a) Relationships among Viewers, Buzz, and Piracy (b) Relationships among TV_rating, Buzz, and Piracy

Piracy − 0.021 (0.010)* Piracy − 0.010 (0.015)
DV: Buzz DV: Buzz
Piracy 3.294 (0.557)*** Piracy 3.294 (0.557)***
Buzz 0.002 (0.001)*** Buzz 0.003 (0.001)***
Piracy − 0.027 (0.010)** Piracy − 0.021 (0.016)
(c) Relationships among Viewership, Buzz, and Piracy (d) Relationships among Viewership, Buzz, and Piracy_new
DV: Viewership DV: Viewership
Piracy − 0.097 (0.080) Piracy_new − 0.005 (0.007)
DV: Buzz DV: Buzz
Piracy 3.294 (0.557)*** Piracy_new 0.310 (0.048)***
DV: Viewership DV: Viewership
Buzz 0.017 (0.004)*** Buzz 0.016 (0.004)***
Piracy − 0.156 (0.081)† Piracy_new − 0.010 (0.007)
27

(a) Relationships among Viewers, Buzz, and Piracy (b) Relationships among TV_rating, Buzz, and Piracy

Note: †p < 0.1.

*
p < 0.05.
**
p < 0.01.
***
Table A14 Table A17

Program list (supplementary sample). Analysis results with search trends (Supplementary Sample)*.
Title Broadcaster Broadcasting date (a) Relationships among TV_rating, (b) Relationships among TV_rating,
Google, and Piracy (Seeds) Google, and Piracy (Leeches)
60 Days, Designated TVN 7/9, 7/15, 7/22, 7/29, 8/5, 8/12, 8/19
Survivor DV: TV_rating DV: TV_rating
A Lover of 3:00 PM on Channel A 7/5, 7/12, 7/19, 7/26, 8/2, 8/9, 8/16 Piracy − 0.006 (0.008) Piracy 0.001 (0.007)
Weekdays Controls Yes Controls Yes
Abyss TVN 6/3, 6/10, 6/17, 6/24 DV: Google DV: Google
Angel’s Last Mission: KBS2 6/5, 6/12, 6/19, 6/26, 7/3, 7/10 Piracy 0.072 (0.016)*** Piracy 0.094 (0.015)***
Love Controls Yes Controls Yes
Arthdal Chronicles TVN 6/1, 6/8, 6/15, 6/22, 6/29, 7/6 DV: TV_rating DV: TV_rating
Be Melodramatic JTBC 8/16, 8/23, 8/30, 9/6 Google 0.050 (0.018)*** Google 0.049 (0.017)**
Class of Lies OCN 7/17, 7/24, 7/31, 8/7, 8/14, 8/21, 8/28, Piracy − 0.008 (0.009) Piracy − 0.002 (0.008)
9/4 Controls Yes Controls Yes
Different Dreams MBC 6/8, 6/15, 6/22, 6/29, 7/6, 7/13 Indirect effect 0.004 Indirect effect 0.005
Elegant House MBN 8/21, 8/28 Direct effect − 0.008 Direct effect − 0.002
Hotel Del Luna TVN 7/13, 7/20, 7/27, 8/3, 8/10, 8/17, 8/24, Total effect − 0.004 Total effect 0.002
8/31 Observations 1077 Observations 1077
Justice KBS2 7/17, 7/24, 7/31, 8/7, 8/14
Let Me Hear Your KBS2 8/5, 8/12, 8/19, 8/26, 9/2 Note: †p < 0.1.
*
Song p < 0.05.
**
Level Up MBN 7/10, 7/17, 7/24, 7/31, 8/7, 8/14 p < 0.01.
***
Mother of Mine KBS2 6/1, 6/8, 6/15, 6/22, 6/29, 7/6, 7/13, 7/ p < 0.001, Two-tailed tests; Standard errors are in parentheses; Control
20, 7/27, 8/3, 8/10, 8/17, 8/24, 8/31 variables include the dummies indicating the dramas, day of the week, broad
My Absolute SBS 6/5, 6/12, 6/19, 6/26, 7/3, 7/10 casting day, broadcasting week, and broadcaster.
Boyfriend
No Work KBS2 6/11, 6/18, 7/2
One Spring Night MBC 6/5, 6/12, 6/19, 6/26, 7/3, 7/10 References
Partners for Justice 2 MBC 6/3, 6/10, 6/17, 6/24, 7/1, 7/8, 7/15, 7/
22 [1] L. Aguiar, B. Martens, Digital music consumption on the internet: evidence from
Perfume KBS2 6/3, 6/10, 6/17, 6/24, 7/1, 7/8, 7/15, 7/ clickstream data, Inf. Econ. Policy 34 (2016) 27–43, https://doi.org/10.1016/j.
22 infoecopol.2016.01.003.
Save Me 2 OCN 6/5, 6/12, 6/19, 6/26 [2] S. Asur, B.A. Huberman, Predicting the future with social media, in: Proceedings of
The Secret Life of My SBS 6/10, 6/17, 6/24 the 2010 IEEE/WIC/ACM International Conference on Web Intelligence and
Secretary Intelligent Agent Technology, Los Alamitos, CA, 2010, pp. 492–499.
[3] B. Andersen, M. Frenz, Don’t blame the P2P file-sharers: the impact of free music
The Wind Blows JTBC 6/3, 6/10, 6/17, 6/24, 7/1, 7/8, 7/15
downloads on the purchase of music CDs in Canada, J. Evol. Econ. 20 (5) (2010)
Voice 3 OCN 6/1, 6/8, 6/15, 6/22, 6/29
715–740.
Watcher OCN 7/6, 7/13, 7/20, 7/27, 8/3, 8/10, 8/17, 8/ [4] Birel, R. (2019). Global piracy hits 190 billion visits in 2018. BroadbandTVnews,
24 Available at: www.broadbandtvnews.com/2019/03/23/global-piracy-hits-190-
Welcome 2 Life MBC 8/5, 8/12, 8/19, 8/26, 9/2 billion-visits-in-2018 (Accessed on Mar 15, 2021).
WWW TVN 6/5, 6/12, 6/19, 6/26, 7/3, 7/10, 7/17, 7/ [5] A.L. Brown, C.F. Camerer, D. Lovallo, To review or not to review? Limited strategic
24 thinking at the movie box office, Am. Econ. J. Microecon. 4 (2) (2012) 1–26.
[6] M.J. Culnan, P.J. McHugh, J.I. Zubillaga, How large US companies can use twitter
and other social media to gain business value, MIS Q. Exec. 9 (4) (2010) 243–259.
[7] B. Danaher, M.D. Smith, Gone in 60 Seconds: the impact of the megaupload
Table A15 shutdown on movie sales, Int. J. Ind Organiz. 33 (1) (2014) 1–8.
Descriptive statistics (supplementary sample). [8] S. Dewan, J. Ramaprasad, Social media, traditional media, and music sales, MIS Q.
38 (1) (2014) 101–121.
Variable Mean Std. Dev. Min Max [9] V. Dhar, E.A. Chang, Does chatter matter? The impact of user-generated content on
music sales, J. Interact. Mark. 23 (4) (2009) 300–307.
TV_rating 5.751 5.938 0.8 27.8 [10] R.Y. Du, Y. Hu, S. Damangir, Leveraging trends in online searches for product
Seeds 495.312 586.812 8 4177 features in market response modeling, J. Mark. 79 (1) (2015) 29–43.
Leeches 63.499 105.786 0 824 [11] D. Godes, D. Mayzlin, Using online conversations to study word-of-mouth
Google 3.919 9.505 0 100 communication, Mark. Sci. 23 (4) (2004) 545–560.
[12] M. Godinho de Matos, P. Ferreira, M.D. Smith, The effect of subscription video-on-
demand on piracy: evidence from a household-level randomized experiment,
Table A16 Manag. Sci. 64 (12) (2017) 5610–5630.
[14] M.J. Hashim, K.N. Kannan, D.T. Wegener, Central role of moral obligations in
Correlations (supplementary sample).
determining intentions to engage in digital piracy, J. Manag. Inf. Syst. 35 (3)
(1) (2) (3) (4) (2018) 934–963.
[15] T. Hennig-Thurau, V. Henning, H. Sattler, Consumer file sharing of motion
TV_rating 1.000 pictures, J. Mark. 71 (4) (2007) 1–18.
Seeds − 0.055 1.000 [16] T. Hennig-Thurau, C. Wiertz, F. Feldhaus, Does Twitter matter? The impact of
Leeches − 0.041 0.878 1.000 microblogging word of mouth on consumers’ adoption of new movies, J. Acad.
Google 0.059 0.322 0.236 1.000 Mark. Sci. 43 (3) (2015) 375–394.
28
[17] S. Hinduja, J.R. Ingram, Social learning theory and music piracy: the differential [38] C. Peukert, J. Claussen, T. Kretschmer, Piracy and box office movie revenues:
role of online and offline peer influences, Crim. Justice Stud. 22 (4) (2009) evidence from megaupload, Int. J. Ind Organiz. 52 (2017) 188–215, https://doi.
405–420. org/10.1016/j.ijindorg.2016.12.006.
[18] Y.L. Hwong, C. Oliver, M. Van Kranendonk, C. Sammut, Y. Seroussi, What makes [39] Y. Qian, Counterfeiters: foes or friends? How counterfeits affect sales by product
you tick? The psychology of social media engagement in space science quality tier, Manag. Sci. 60 (10) (2014) 2381–2400.
communication, Comput. Hum. Behav. 68 (2017) 480–492. [40] H. Rui, Y. Liu, A. Whinston, Whose and what chatter matters? The effect of tweets
[19] Kain, E. (2012). A brief history of US anti-piracy efforts. Forbes, www.forbes.com/ on movie sales, Decis. Support Syst. 55 (4) (2013) 863–870.
sites/erikkain/2012/02/15/a-brief-history-of-us-anti-piracy-efforts/ [41] A. Schwabach, Internet and the Law: Technology, Society, and Compromises, ABC-
#4198c491779b (Accessed on Aug. 15, 2020). CLIO, Santa Barbara, CA, 2014.
[20] E.V. Karniouchina, C. Uslay, G. Erenburg, Do marketing media have life cycles? [42] Smith, G. (2018). Blame netflix, disney, and the dodgers for killing Americans’
The case of product placement in movies, J. Mark. 75 (3) (2011) 27–48. Cable-TV bundle. Los Angeles Times, https://www.latimes.com/business/la-fi-who-
[21] Koschmann, A., & Qian, Y. (2020). Latent estimation of piracy quality and its effect killed-cable-tv-bundle-20180816-story.html (Accessed on Aug 15, 2020).
on revenues and distribution: the case of motion pictures, Natl. Bur. Econ. Res., [43] M.D. Smith, R. Telang, Competing with free: the impact of movie broadcasts on
Available at www.nber.org/papers/w27649. DVD sales and internet piracy, MIS Q. 33 (2) (2009) 321–338.
[22] T. Kretschmer, C. Peukert, Video killed the radio star? Online music videos and [44] R. Telang, J. Waldfogel, Piracy and new product creation: a bollywood story, Inf.
recorded music sales, Inf. Syst. Res. 31 (3) (2020) 776–800. Econ. Policy 43 (2018) 1–11.
[23] J.L. Krull, D.P. MacKinnon, Multilevel modeling of individual and group level [45] Thier, D. (2013). This ’Game of Thrones’ director isn’t bothered by piracy. Forbes,
mediated effects, Multivar. Behav. Res. 36 (2) (2001) 249–277. Available at: www.forbes.com/sites/davidthier/2013/02/26/this-game-of-thron
[24] S.J. Liebowitz, Research note—testing file sharing’s impact on music album sales es-director-isnt-bothered-by-piracy/?sh=624a5af14b76.
in cities, Manag. Sci. 54 (4) (2008) 852–859. [46] Van der Sar, E. (2020). Game Developer Sees Boost in Sales After Releasing Official
[25] S.J. Liebowitz, The Oberholzer-Gee/Strumpf File-sharing Instrument Fails the Torrent, Torrentfreak, Available at: torrentfreak.com/game-developer-sees-boost-i
Laugh Test, The University of Texas at Dallas, 2010. https://papers.ssrn.com/soL3/ n-sales-after-releasing-official-torrent-200113.
papers.cfm?abstract_id=1598037. Accessed on Aug 15, 2020. [47] C.J. Vargo, Toward a tweet typology: contributory consumer engagement with
[26] Liu, Z. (2018). Quantifying the heterogeneous effects of piracy on the demand for brand messages by content type, J. Interact. Adv. 16 (2) (2016) 157–168.
movies. Working Paper, Available at publish.uwo.ca/~zliu328/jobmarketpaper. [48] Z. Yang, J. Wang, M. Mourali, Effect of peer influence on unauthorized music
pdf. downloading and sharing: the moderating role of self-construal, J. Bus. Res. 68 (3)
[27] S. Lu, X. Wang, N. Bendle, Does piracy create online word of mouth? An empirical (2015) 516–525.
analysis in the movie industry, Manag. Sci. 66 (5) (2020) 2140–2162. [49] L. Zhang, Intellectual property strategy and the long tail: evidence from the
[28] L. Ma, A.L. Montgomery, P.V. Singh, M.D. Smith, An empirical analysis of the recorded music industry, Manag. Sci. 64 (1) (2018) 24–42.
impact of pre-release movie piracy on box office revenue, Inf. Syst. Res. 25 (3)
(2014) 590–603.
Dongyeon Kim is an assistant professor at the Department of Business Administration,
[29] Ma, L., Montgomery, A., & Smith, M.D. (2016). The Dual impact of movie piracy
Catholic University of Korea. He earned his Ph.D. in Management Engineering from Korea
on box-office revenue: cannibalization and promotion. Working Paper, Available at
Advanced Institute of Science and Technology (KAIST). His research interests include the
SSRN 2736946.
business value of data analytics and the mechanisms of change under advanced technol
[30] E. Martikainen, Does file-sharing reduce DVD sales? NETNOMICS: Econ. Res.
ogy. His prior research has been published in International Journal of Electronic Commerce,
Electron. Netw. 15 (1) (2014) 9–31.
Electronic Commerce Research and Applications, and several other journals.
[31] Mcclintock, P. (2021) Theater owners sound piracy alarm Over Disney+ ‘Black
Widow’ streaming reveal, Available at: www.hollywoodreporter.com/business/
business-news/theater-owners-sound-piracy-alarm-over-disney-black-widow-stre Kyuhong Park is an assistant professor at the Department of Business Administration,
aming-reveal-1234985460. Inha University. He earned his Ph.D. in Management Engineering from Korea Advanced
[32] J. McKenzie, Graduated response policies to digital piracy: do they increase box Institute of Science and Technology (KAIST). His research interests include Internet of
office revenues of movies? Inf. Econ. Policy 38 (2017) 1–11. 10.1016/j.infoecopol. Things, digital piracy, and information privacy. His prior research has been published in
2016.12.004. International Journal of Electronic Commerce, Computers in Human Behavior, Telematics and
[33] Melancon, J.M. (2020). Research shows post-release digital movie piracy can boost Informatics, and several other journals.
ticket sales, Phys Org, https://phys.org/news/2020-08-post-release-digital-movie-
piracy-boost.html.
Youngsok Bang is an assistant professor at the School of Business, Yonsei University.
[34] Mention, The Twitter engagement report 2018. Mention, 2018. https://info.men
Before joining Yonsei University, he spent four years at the Chinese University of Hong
tion.com/hubfs/Twitter%20Engagement%20Report%202018%20%7C%20Men
Kong as an assistant professor in the Department of Decision Sciences and Managerial
tion.pdf. Accessed on Aug 15, 2020.
Economics. He received his B.S., M.S., and Ph.D. in Management Engineering from Korea
[35] F. Oberholzer-Gee, K. Strumpf, The effect of file sharing on record sales: an
Advanced Institute of Science and Technology (KAIST) and his second Ph.D. in Manage
empirical analysis, J. Polit. Econ. 115 (1) (2007) 1–42.
ment (Concentration: Information Systems) from McGill University. His current interests
[36] A.G. Peace, D.F. Galletta, J.Y. Thong, Software piracy in the workplace: a model
lie in the business value of data analytics, especially in the areas of mobile business and
and empirical test, J. Manag. Inf. Syst. 20 (1) (2003) 153–177.
FinTech. His research has appeared in Information Systems Research, Journal of MIS, MIS
[37] Pearce, J. (2020). Piracy surges during coronavirus lockdown. IBC, Available at
Quarterly, and several other journals.
www.ibc.org/piracy-surges-during-coronavirus-lockdown/5820.article (Accessed
on Mar 15, 2021).
29

Effect of TV Drama Piracy

Uploaded by

Copyright:

Available Formats

Effect of TV Drama Piracy

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Effect of TV Drama Piracy

Uploaded by

Copyright:

Available Formats

Information & Management 59 (2022) 103599

Contents lists available at ScienceDirect

Information & Management

The effect of TV drama piracy: An analysis of digital piracy users, internet

1. Introduction word-of-mouth, the consumption boosted by digital piracy can promote

Table 6(a) shows the estimates of the relationships among Viewers,

DV: Viewers DV: TV_rating DV: Viewers

Note: †p < 0.1.

DV: Viewers DV: TV_rating

(c) Relationships among Viewers, (d) Relationships among TV_rating,

Note: †p < 0.1.

from dramas. Given that ’Moviest,’ ’Pornt,’ ’Softwaret,’ and ’Gamest’

DV: Viewers DV: TV_rating DV: Viewers

Note: †p < 0.1.

DV: Viewers DV: TV_rating

Total effect 0.006 Total effect 0.019

Note: †p < 0.1.

DV: Viewers DV: TV_rating DV: Viewers

Note: †p < 0.1.

Table 12 Table 13 (continued )

(c) Relationships among Viewers, (d) Relationships among TV_rating, Table 14

Table 15 Table 16 (continued )

(c) Relationships among Viewers, (d) Relationships among TV_rating,

(c) Relationships among Viewers, (d) Relationships among TV_rating,

Tracker (log) 2.175 1.114 0 4.444

promotional effects of pirated content. Along with this research stream,

DV: Viewers DV: TV_rating DV: Viewers

Note: †p < 0.1.

CRediT authorship contribution statement Acknowledgments

Declarations of Competing Interest

Appendix 1. Results of PVAR analysis

• Piracy, Viewer, and Google

Selection order criteria

lag CD J J p-value MBIC MAIC MQIC

• Piracy, TV_rating, and Google

Selection order criteria

lag CD J J p-value MBIC MAIC MQIC

• Piracy, Viewer, and YouTube

Selection order criteria

lag CD J J p-value MBIC MAIC MQIC

• Piracy, TV_rating, and YouTube

Selection order criteria

Lag CD J J p-value MBIC MAIC MQIC

Panel VAR-Granger causality Wald test

• Piracy, Viewer, and Tweets

Selection order criteria

lag CD J J p-value MBIC MAIC MQIC

• Piracy, TV_rating, and Tweets

Selection order criteria

lag CD J J pvalue MBIC MAIC MQIC

• Piracy, Viewer, and Engagements

Selection order criteria

lag CD J J p-value MBIC MAIC MQIC

• Piracy, TV_rating, and Engagements

Selection order criteria

lag CD J J p-value MBIC MAIC MQIC

• Piracy, Viewer, and Interactions

Selection order criteria

Lag CD J J p-value MBIC MAIC MQIC